CN108446374A - User view prediction technique, device, electronic equipment, storage medium - Google Patents
User view prediction technique, device, electronic equipment, storage medium Download PDFInfo
- Publication number
- CN108446374A CN108446374A CN201810218834.8A CN201810218834A CN108446374A CN 108446374 A CN108446374 A CN 108446374A CN 201810218834 A CN201810218834 A CN 201810218834A CN 108446374 A CN108446374 A CN 108446374A
- Authority
- CN
- China
- Prior art keywords
- user
- behavior
- historical
- view prediction
- history
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/955—Retrieval from the web using information identifiers, e.g. uniform resource locators [URL]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9537—Spatial or temporal dependent retrieval, e.g. spatiotemporal queries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
A kind of user view prediction technique of present invention offer, device, electronic equipment, storage medium, user view prediction technique include:User's history behavior sequence is acquired, the user's history behavior sequence includes multiple historical user's behavior items, and each historical user's behavior item includes historical user's behavior and executes the location information of historical user's behavior;A user view prediction model is trained according to the user's history behavior sequence;The real-time behavior sequence of user is acquired, the real-time behavior sequence of user includes multiple active user behavior items, and each active user behavior item includes active user behavior and executes the location information of active user behavior;And according to the real-time behavior sequence of the user, user behavior is predicted by the user view prediction model.Method and device provided by the invention can preferably depict the Demand perference of user.
Description
Technical field
The present invention relates to computer application technologies more particularly to a kind of user view prediction technique, device, electronics to set
Standby, storage medium.
Background technology
In order to meet user individual interactive experience, to improve user activity and user's viscosity, service/goods provider
The especially needed concern promptness of interaction design and content information of application program or webpage, correlation, customization etc..For
Realize correlation and the customization of interaction design and content information, it is necessary first to excavate user demand, understand user view.Mesh
Before, there are two types of the technical solutions of excavating user view:One is user's portrait, need by mass data with to user into rower
Labelization, and then judge by label the interest and preference of user;Another kind is historical behavior statistical value, based on artificial rule
Probability statistics are carried out to user data, thus judge the interest and preference of user.
However, the either mode of label or statistical value, due to its discrete data type, it is difficult to realize fine and smooth use
Family is distinguished, and then is obtained careful user preference and portrayed.
Invention content
The present invention provides a kind of user view prediction technique, device, electricity to overcome defect existing for above-mentioned the relevant technologies
Sub- equipment, storage medium, and then one is overcome caused by the limitation and defect of the relevant technologies at least to a certain extent
Or multiple problems.
According to an aspect of the present invention, a kind of user view prediction technique is provided, including:
User's history behavior sequence is acquired, the user's history behavior sequence includes multiple historical user's behavior items, each institute
Historical user's behavior item is stated to include historical user's behavior and execute the location information of historical user's behavior;
A user view prediction model is trained according to the user's history behavior sequence;
The real-time behavior sequence of user is acquired, the real-time behavior sequence of user includes multiple active user behavior items, each institute
Active user behavior item is stated to include active user behavior and execute the location information of active user behavior;And
According to the real-time behavior sequence of the user, user behavior is predicted by the user view prediction model.
According to another aspect of the invention, a kind of user view prediction meanss are also provided, including:
First acquisition module, for acquiring user's history behavior sequence, the user's history behavior sequence includes multiple goes through
History user behavior item, each historical user's behavior item include historical user's behavior and execute the position letter of historical user's behavior
Breath;
Training module, for training a user view prediction model according to the user's history behavior sequence;
Second acquisition module, for acquiring the real-time behavior sequence of user, the real-time behavior sequence of user includes multiple realities
When user behavior item, each active user behavior item include active user behavior and execute active user behavior position letter
Breath;And
Prediction module, for according to the real-time behavior sequence of the user, predicting to use by the user view prediction model
Family behavior.
According to another aspect of the invention, a kind of electronic equipment is also provided, the electronic equipment includes:Processor;Storage
Medium, is stored thereon with computer program, and the computer program executes step as described above when being run by the processor.
According to another aspect of the invention, a kind of storage medium is also provided, computer journey is stored on the storage medium
Sequence, the computer program execute step as described above when being run by processor.
Compared with prior art, advantage of the invention is that:
In view of the different location of user behavior will produce the variation of user view, the present invention is in user's history behavior sequence
And user behavior is made to increase the attribute of location information when in the real-time behavior sequence of user, thereby, it is possible to dynamically track user's row
For when geographical location transfer and variation, via the study of user view prediction model, with according to active user behavior and position
Information prediction user behavior is set, so as to preferably depict the Demand perference of user based on location information.
Description of the drawings
Its example embodiment is described in detail by referring to accompanying drawing, above and other feature of the invention and advantage will become
It is more obvious.
Fig. 1 shows the flow chart of user view prediction technique according to the ... of the embodiment of the present invention.
Fig. 2 shows the flow charts of user's history behavior sequence vectorization according to the ... of the embodiment of the present invention.
Fig. 3 shows the flow chart of determining mode input and output according to the ... of the embodiment of the present invention.
Fig. 4 shows the flow chart of the real-time behavior sequence vectorization of user according to the ... of the embodiment of the present invention.
Fig. 5 shows a kind of schematic diagram of user view prediction model according to the ... of the embodiment of the present invention.
Fig. 6 shows the block diagram of user view prediction meanss according to the ... of the embodiment of the present invention.
Fig. 7 shows the block diagram of the user view prediction meanss according to a specific embodiment of the invention.
Fig. 8 schematically shows a kind of computer readable storage medium schematic diagram in exemplary embodiment of the present.
Fig. 9 schematically shows a kind of electronic equipment schematic diagram in exemplary embodiment of the present.
Specific implementation mode
Example embodiment is described more fully with reference to the drawings.However, example embodiment can be with a variety of shapes
Formula is implemented, and is not understood as limited to example set forth herein;On the contrary, thesing embodiments are provided so that the present invention will more
Fully and completely, and by the design of example embodiment comprehensively it is communicated to those skilled in the art.Described feature, knot
Structure or characteristic can be in any suitable manner incorporated in one or more embodiments.
In addition, attached drawing is only the schematic illustrations of the present invention, it is not necessarily drawn to scale.Identical attached drawing mark in figure
Note indicates same or similar part, thus will omit repetition thereof.Some block diagrams shown in attached drawing are work(
Energy entity, not necessarily must be corresponding with physically or logically independent entity.Software form may be used to realize these work(
Energy entity, or these functional entitys are realized in one or more hardware modules or integrated circuit, or at heterogeneous networks and/or place
These functional entitys are realized in reason device device and/or microcontroller device.
Flow chart shown in attached drawing is merely illustrative, it is not necessary to including all steps.For example, the step of having
It can also decompose, and the step of having can merge or part merges, therefore, the sequence actually executed is possible to according to actual conditions
Change.
Fig. 1 diagrammatically illustrates the flow chart of user view prediction technique according to the ... of the embodiment of the present invention.With reference to figure 1, institute
The user view prediction technique stated may comprise steps of:
Step S110:User's history behavior sequence is acquired, the user's history behavior sequence includes multiple historical user's rows
For item, each historical user's behavior item includes historical user's behavior and executes the location information of historical user's behavior;
Step S120:A user view prediction model is trained according to the user's history behavior sequence;
Step S130:The real-time behavior sequence of user is acquired, the real-time behavior sequence of user includes multiple active user rows
For item, each active user behavior item includes active user behavior and executes the location information of active user behavior;And
Step S140:According to the real-time behavior sequence of the user, user's row is predicted by the user view prediction model
For.
In the user view prediction technique of exemplary embodiments of the present invention, it is contemplated that the different location of user behavior
The variation of user view is will produce, makes user behavior when the present invention is in user's history behavior sequence and the real-time behavior sequence of user
The attribute for increasing location information, thereby, it is possible to the transfer and variation in geographical location when dynamically tracking user behavior, via with
The study of family Intention Anticipation model, to predict user behavior according to active user behavior and location information, so as to be based on position
Confidence breath preferably depicts the Demand perference of user.
Specifically, with the movement of user, even if identical behavior, the user demand of expression and stresses intensity and be
Different.For example, user equally implements search chafing dish this behavior at home and in popular commercial circle, user may be partially at home
To merchandise news diversification, information browse acquisition is laid particular emphasis on, user may lay particular emphasis on and consume nearby in commercial circle.Lead in the present invention
It crosses behavior and the combination of geographical location attribute, can learn and capture this category information by neural network model, to more preferable
The Demand perference for depicting user.
In various embodiments of the present invention, the real-time behavior sequence of user has category identical with user's history behavior sequence
Property and structure;Active user behavior item has attribute identical with historical user's behavior item and structure;Active user behavior has
Attribute identical with historical user's behavior and structure.The real-time behavior sequence of user is different only in that with user's history behavior sequence
Active user behavior in the real-time behavior sequence of user occurs in real time, and historical user's row in user's history behavior sequence
It is to occur in the time previous.
The user view prediction technique of exemplary embodiments of the present invention will be illustrated below.
Above-mentioned steps S110 acquisition user's history behavior sequences may further include following steps:
First, multiple historical user's behavior items in a historical time section are acquired.
Specifically, the historical time section, for example, can be " one day " before current time, " 12 hours ",
" 6 hours ", " 1 hour " etc., the present invention are not so limited.The present invention is by acquiring multiple history in a historical time section
User behavior item, and be trained with this, to realize that the short-term real-time intention of user is portrayed, to improve user's behavior prediction
Precision and accuracy.
Historical user's behavior item includes historical user's behavior and executes the location information of historical user's behavior.History
User behavior can be search key, screening category, exposure commodity (such as sharing, forward the behavior of commodity), click commodity,
Lower list commodity etc., the present invention is not so limited.It is appreciated that in historical user's behavior, search key and screening category
To convert unrelated user behavior with commodity;It is to convert related user with commodity to expose commodity, click commodity and lower single commodity
Behavior.In various embodiments of the present invention, it can be used unrelated historical user's behavior is converted with commodity as first kind history
Family behavior will convert relevant historical user's behavior as second class historical user's behavior with commodity.In some specific embodiments
In, the form that character string may be used indicates historical user's behavior.For example, " the red candles of q ", sieve can be converted to by searching for red candle
Beauty competition foodstuff, which can be converted to " s cuisines ", search paddy taste can mostly be converted to " q paddy tastes are more ", click commodity A to be converted to
" cp553203 " (wherein, commodity A is the interested commodity of user, and 553203 be the ID of commodity A), lower list commodity A can be converted
For " op553203 ".For the sake of above-mentioned character string forms are only schematic, the present invention is not so limited, and the present invention may be used
Different character string conversion regime, to realize the string representation of historical user's behavior.
The location information of execution historical user's behavior for example can be when executing historical user's behavior, where user
Longitude and latitude.Location information can be according to any acquisition in such as under type:Acquisition executes the user terminal of historical user's behavior
Locating module location information;Convert the addresses ip of user terminal to location information;According to user terminal and communication base station
The determining location information etc. of interaction.User terminal is not limited to smart mobile phone, tablet computer, desktop computer, electricity on knee
Brain etc..In certain embodiments, the form that character string may be used indicates location information.Specifically, position can be made
Information obtains string representation by geohash algorithms, for example, the string representation wwgq3y of location information, wherein ww are indicated
City rank size geographic range, wwgq3y can indicate rank size geographic range in commercial circle under city.Above-mentioned character string forms
Only schematically for the sake of, the present invention it is not so limited, different character string conversion regimes may be used in the present invention, with realize
The string representation of location information.
As a result, in one particular embodiment of the present invention, historical user's behavior item is with Ai=<geoi,acti>Form
It indicates, wherein acti is the string representation of historical user's behavior, and geoi is the word for the location information for executing historical user's behavior
Symbol string indicates.For example, historical user's behavior item can be Ai=<The red candle of wwgq3y, q>, wherein historical user's behavior is to search
Suo Hong candles, it is wwgq3y to execute and search for the location information of red candle.
Next, multiple historical user's behavior items in the historical time section to be pressed to the execution time of historical user's behavior
Sequence forms the user's history behavior sequence.
Specifically, in above-mentioned specific embodiment, the user's history behavior sequence of acquisition can be:<wwgq3y,q
Red candle>,<wwgq3y,cp553203>,<wwgq3y,cp98162974>,<Wwgq3y, s cuisines>,<Wwgq3y, q paddy taste are more
>,<wwgq3y,cp5458065>.Wherein, which sorts by the execution sequence of historical user's behavior, that is, goes through
History user behavior executes in the following order:It searches for red candle, click the commodity that ID is 553203, the quotient that click ID is 98162974
Product, screening cuisines category, search paddy taste is more, clicks the commodity that ID is 5458065.
Above-mentioned steps S120 trains a user view prediction model can be further according to the user's history behavior sequence
Include the following steps:
First, make the user's history behavior sequence vectorization, wherein respectively to each in the user's history behavior sequence
Historical user's behavior of historical user's behavior item and the location information carry out vectorization, and by the historical user of vectorization
Behavior and location information splice the user's history behavior item for obtaining vectorization.
Next, training the user view prediction model according to the user's history behavior sequence of vectorization.
At one of above-described embodiment in the specific implementation, referring to Fig. 2, above-mentioned steps S120 is according to the user's history behavior
Sequence trains a user view prediction model to may further include following steps:
Step S211:Historical user's behavior vectorization is made using word2vev algorithms.
Specifically, being appreciated that and identifying vector, common method in order to which the word of historical user's behavior is switched to computer
Be one-hot coding, such as " s cuisines " can vector turn to [0,0,0,1,0,0,0 ..., 0,0,0];" the red candles of q " can vector
Turn to [0,0,0,0,0,1,0 ..., 0,0,0], " cp553203 " can vector turn to [0,0,0,0,0,0,1 ..., 0,0,0], to
In amount only there are one value be 1, remaining all be 0.However, One-hot coding the problem is that, vectorial dimension depend on behavior
Number, the corresponding vector of all behaviors is combined into a matrix, then the matrix can be excessively sparse, causes dimension disaster, and
It is independent from each other between behavior, does not see incidence relation that may be present between behavior.Therefore, the present invention passes through word2vec
Algorithm converts historical user's behavior to low-dimensional, continuous semantic vector, such as " s cuisines " can vector turn to [- 1.16475,
0.283087,-1.1012,0.407279,….,0.908895,-0.544638]100.Above is only schematically to illustrate this
A kind of vectorization mode of historical user's behavior of invention, the present invention are not so limited.
Step S212:Make the vector of position of the character string forms by mapping table.
Specifically, the string representation for location information, indicates different big by intercepting character string different length
Small geographic range, descending such as city, commercial circle equigranular, then it is expressed as real vector by way of tabling look-up.
For example, for wwgq3y (string representation obtained by geohash algorithms), ww is with indicating city rank size
Range is managed, wwgq3y can indicate rank size geographic range in commercial circle under city, and mapping table, which may be used, for wwgq3y tables look-up
Mode, be mapped onto one 32 dimension vector [0,0,0,0,1,0,0,0,0,0,0 ..., 0,0,1,0,0,0]32, wherein to
Preceding 10 Wesy of amount indicates that city granularity, rear 22 dimension table show commercial circle granularity.
Accordingly, for historical user's behavior item Ai, V is obtained after vectorizationi=[Egeo, Eact], wherein Eact be to
Historical user's behavior of quantization, Egeo are the location information of vectorization.
In a specific embodiment, historical user's behavior item<Wwgq3y, s cuisines>Vector it is for example sliceable after indicate
It is as follows:
V=[0,0,0,0,1,0,0,0,0,0,0 ..., 0,0,1,0,0,0, -1.16475,0.283087, -1.1012,
0.407279,…,0.908895,-0.544638]132
It is appreciated that due to historical user's behavior and the string representation of location information difference, divide in the present invention
It is other that different vectorization algorithms is used to them.Since historical user's behavior is higher-dimension, discrete, sparse behavioral data,
Historical user's behavior is made to be converted into low-dimensional, continuous semantic vector historical user's behavior vectorization using word2vev algorithms
While, relationship that can also be between embodiment behavior.Since the string representation of location information is regular, limited number
According to therefore, the operation difficulty that the mode that acquisition mapping table is tabled look-up allows vector of position to reduce vectorial converting algorithm adds
Fast vectorization speed.
Step S213:The user view prediction model is trained according to the user's history behavior sequence of vectorization.
Further, train a user view pre- according to the user's history behavior sequence referring to Fig. 3, above-mentioned steps S120
It surveys model and may further include following steps:
Step S221:Historical user's behavior item is divided into the first kind historical user with first kind historical user's behavior
Behavior item and second class historical user's behavior item with second class historical user's behavior.
It, will be with specifically, unrelated historical user's behavior can be converted as first kind historical user's behavior using with commodity
Commodity convert relevant historical user's behavior as second class historical user's behavior.For example, first kind historical user's row
To may include search key, screening category;Second class historical user's behavior may include exposure commodity, click quotient
Product, lower single commodity.Further, in the embodiment of the string representation of above-mentioned historical user's behavior, red candle is searched for
Character string is " the red candles of q ", the character string of screening cuisines category is " s cuisines ", the character string more than search paddy taste is " q paddy tastes
It is more ", to click the character string of commodity A be " cp553203 ", it will be understood that there are commodity in the string representation of historical user's behavior
ID is second class historical user's behavior, is first kind historical user's behavior without commodity ID.Above-described embodiment is only
For the sake of schematically, the present invention is not so limited.
Step S222:By first kind historical user's behavior item or the first kind historical user behavior item and described second
Input of the class historical user's behavior item as the user view prediction model.
In one embodiment, can only by first kind historical user's behavior item via step S211 as shown in Figure 2 and
Input after S212 progress vectorizations as the user view prediction model.
It In yet another embodiment, can be by first kind historical user's behavior item and the second class historical user behavior item
Via the input as the user view prediction model after step S211 and S212 progress vectorization as shown in Figure 2.
Step S223:Using the second class historical user behavior as the output of the user view prediction model.
Specifically, the output of the user view prediction model can not include location information.In other embodiment
In, when training, the output of the user view prediction model can also include location information, in actual prediction, delete position
Information.Second class historical user's behavior is predicted after word2vev algorithm vectorizations can be used as the user view
The output of model.Further, the input of either user view prediction model still exports, and still presses execution sequence and arranges
Sequence.
In one particular embodiment of the present invention, in the training stage, user view prediction can be divided as follows
Model is output and input:
For, by the multiple historical user's behavior items for executing sequencing sequence, being executed from most end in historical time section
Historical user's behavior item start, judge whether historical user's behavior of historical user's behavior item is second class historical user's row
For if so, using the second class historical user behavior as the output of user view prediction model, and from current historical user
Behavior item is re-executed to previous item judges whether historical user's behavior of historical user's behavior item is the second class historical user
The step of behavior;If it is not, then current historical user's behavior item is started to the historical user's behavior item executed at first as institute
State the input of user view prediction model.For example, user's history behavior sequence is:<The red candle of wwgq3y, q>,<wwgq3y,
cp553203>,<Wwgq3y, s cuisines>,<Wwgq3y, q paddy taste are more>,<wwgq3y,cp5458065>,<wwgq3y,
cp98162974>,<wwgq3y,op98162974>, then from the last<wwgq3y,op98162974>Start to judge its history
Whether user behavior is second class historical user's behavior, and judging result op98162974 is second class historical user's behavior, then will<
wwgq3y,op98162974>Output as model;Again to previous item<wwgq3y,cp98162974>Judge its historical user
Whether behavior is second class historical user's behavior, and judging result cp98162974 is second class historical user's behavior, then will<
wwgq3y,cp98162974>Output as model;Again to previous item<wwgq3y,cp5458065>Judge its historical user's row
Whether to be second class historical user's behavior, judging result cp5458065 is second class historical user's behavior, then will<wwgq3y,
cp5458065>Output as model;Then, then to previous item<Wwgq3y, q paddy taste are more>Judging its historical user's behavior is
No is second class historical user's behavior, and judging result q paddy taste is not mostly second class historical user's behavior, then will<Wwgq3y, q paddy
Taste is more>To the historical user's behavior item executed at first<The red candle of wwgq3y, q>Input as model.Accordingly, for above-mentioned use
Family historical behavior sequence,<The red candle of wwgq3y, q>,<wwgq3y,cp553203>,<Wwgq3y, s cuisines>,<Wwgq3y, q paddy
Taste is more>Input by execution sequence as user view prediction model;Cp5458065, cp98162974, op98162974 are by holding
Output of the row sequence as user view prediction model.In such embodiments, related with commodity conversion in order to finally predict
Historical user's behavior, therefore, using the second class historical user behavior as the output of user view prediction model.However, with
In the historical behavior sequence of family, it will usually the case where there are first kind historical user behavior and the second class user behavior intervals, in order to
Make user's Intention Anticipation model there are enough inputs to obtain more accurate prediction, therefore, uses in the present embodiment as above
The mode of judgement is with since the most end of user's historical behavior sequence, by the historical user of first first kind historical user's behavior
Input of the behavior item to the historical user's behavior item executed at first as user view prediction model.
In a specific implementation mode of the embodiment, it is that boundary will be in historical time section that can descend the behavior of single commodity
Be divided into multiple user's history behavior sequences by the multiple historical user's behavior items for executing sequencing sequence, and each user goes through
Adjacent historical user's behavior interval in history behavior sequence is less than scheduled time threshold value (such as 5 minutes, 10 minutes, 20 minutes
Deng).For example, when being divided into 60 minutes between the red candles of adjacent historical user's behavior q and cp553203 in user's history behavior sequence, by
Therebetween every overlong time, it will be understood that the different wishes that above-mentioned two historical user's behaviors are based largely on user are come
It carries out, therefore, can incite somebody to action<The red candle of wwgq3y, q>With<wwgq3y,cp553203>It is respectively divided to two user's history behavior sequences
In row.Can only the corelation behaviour sequence of lower single commodity be learnt and be analyzed as a result, at the same by the setting of time threshold come
Primarily determine the correlation of historical user's behavior and lower single commodity behavior.It is formed and is used by the stronger historical user's behavior of correlation
Family historical behavior sequence, and then improve the accuracy of user view prediction model.
Further, the user view prediction model is a Recognition with Recurrent Neural Network RNN, basis described in above-mentioned steps S120
It further includes following steps that the user's history behavior sequence, which trains a user view prediction model,:
Using at least partly described user's history behavior sequence as the input of the Recognition with Recurrent Neural Network RNN.
Using at least partly described user's history behavior sequence as the output of the Recognition with Recurrent Neural Network RNN, wherein described
The output of Recognition with Recurrent Neural Network RNN is converted according to state vector and is obtained, and the state vector is according to the Recognition with Recurrent Neural Network RNN
Input conversion obtain.
Conversions and the state vector of the training Recognition with Recurrent Neural Network RNN to the state vector are refreshing to the cycle
The conversion of output through network RNN.
The structure and training method of specific Recognition with Recurrent Neural Network RNN is illustrated hereinafter in conjunction with Fig. 5, herein not
It gives and repeating.
Above-mentioned steps S140 predicts user according to the real-time behavior sequence of the user, by the user view prediction model
Behavior may further include following steps:
First, make the real-time behavior sequence vectorization of the user, wherein respectively to each in the real-time behavior sequence of the user
The active user behavior of active user behavior item and the location information carry out vectorization, and by the active user of vectorization
Behavior and location information splice the real-time behavior item of user for obtaining vectorization.
Next, according to the real-time behavior sequence of the user of vectorization, predicted by the user view prediction model
User behavior.
At one of above-described embodiment in the specific implementation, referring to Fig. 4, above-mentioned steps S140 is according to the real-time behavior of the user
Sequence predicts that user behavior may further include following steps by the user view prediction model:
Step S411:Active user behavior vectorization is made using word2vev algorithms.
Specifically, the vectorization algorithm of active user behavior is identical as the vectorization algorithm of historical user's behavior.
Step S412:Make the vector of position of the character string forms by mapping table.
Specifically, step S412 can be identical as step S212.
Step S413:It is pre- by the user view prediction model according to the real-time behavior sequence of the user of vectorization
Survey user behavior.
Specifically, in a specific embodiment of the present invention, the user behavior of user view prediction model prediction can be
Relevant user behavior is converted with commodity.
It should be noted that although describing each step of method in the present invention with particular order in the accompanying drawings, this is simultaneously
Undesired or hint must execute these steps according to the particular order, or have to carry out the step ability shown in whole
Realize desired result.Additional or alternative, it is convenient to omit multiple steps are merged into a step and executed by certain steps,
And/or a step is decomposed into execution of multiple steps etc..
Fig. 5 shows a kind of schematic diagram of user view prediction model according to the ... of the embodiment of the present invention.It is provided by the present invention
User view prediction model can be Recognition with Recurrent Neural Network RNN.
Specifically, the present invention uses the Seq2Seq models of Recognition with Recurrent Neural Network RNN.In Figure 5, n, m be more than or equal to
1 positive integer.Seq2Seq models include three parts, i.e. Encoder330 (coding layer), Decoder350 (decoding layer) and
Connect the intermediate state vector 340 of Encoder330 and Decoder350.
At least partly historical user's behavior item A in above-mentioned user's history behavior sequence1To AnThrough 320 vectorization of vectorization layer
For V1To Vn, V1To VnSeq2Seq models are inputted, in Encoder330:
h1=f (UV1+W h0+b);
h2=f (UV2+W h1+b);
h3=f (UV3+W h2+b);
...
hn=f (UVn+W hn-1+ b),
Wherein, U, W, b are in h1To hnIt is identical in the calculating of (the first hidden vector), and backpropagation at any time can be passed through
Algorithm BPTT carrys out the gradient of counting loss function, to which iteration updates U, W, b.h0Can be that preset constant can also be with
U, iteration updates together by W, b.
The calculation of state vector C340 can be following any middle mode:
C=hn;
C=q (hn);
C=q (h1,h2,h3...,hn)。
Wherein, above-mentioned C=q (hn) and C=q (h1,h2,h3...,hn) in q refer to the functional relation of pre-determining.It changes
Yan Zhi, Encoder330 by learn input, be encoded into the state vector 340 of a fixed size, then by state to
Amount 340 is transmitted to Decoder350.Decoder350 is by exporting the study of state vector 340.
In the embodiment shown in fig. 5, in Decoder350, h can sequentially be calculated according to state vector C3401' to hm’
(the second hidden vector), Y1To YmRespectively according to h1' to hm' calculate acquisition.
Specifically, inputs of the Encoder330 in j-th of Recognition with Recurrent Neural Network RNN of jth stage reception, and according to
The first hidden vector h that Encoder330 is exported in -1 stage of jthj-1Calculate Encoder330 the jth stage export first it is hidden to
Measure hj, j is the integer for being less than or equal to n more than or equal to 1.Decoder350 receives the state vector 340, and root in the i-th stage
The the second hidden vector h exported in the (i-1)-th stage according to Decoder350i-1' calculate Decoder350 the i-th stage export second
Hidden vector hi', and it is based on the described second hidden vector hi' provide i-th of Recognition with Recurrent Neural Network RNN output, i be more than etc.
It is less than or equal to m in 1.
Form the defeated of the Seq2Seq models of Recognition with Recurrent Neural Network RNN through vectorization according to user's history behavior sequence as a result,
Enter and export and solution training is carried out come the parameter of the Seq2Seq models to Recognition with Recurrent Neural Network RNN with this.When by the real-time row of user
It is sequence after vectorization is as the input of the Seq2Seq models of Recognition with Recurrent Neural Network RNN, Recognition with Recurrent Neural Network can be obtained
The output of the Seq2Seq models of RNN.Specifically, by after the output inverse quantization obtained, you can obtain user's row of prediction
For.
Specifically, each in the Encoder330 and Decoder350 of the Seq2Seq models of Recognition with Recurrent Neural Network RNN
Unit (h1To hn, h1' to hm') can be LSTM (Long Short-Term Memory, shot and long term memory network) or GRU
The structure of (Gated Recurrent Unit, gating cycle unit).Encoder330 is by input coding at fixed size state
Vector 340, this process is actually an information lossy compression process, is increased with the length of list entries, this conversion
The process of vector is bigger to the loss of information.The state vector 340 of fixed size is so that Decoder350 can not directly go to close
Note the more details of input information.Based on this defect, the present invention can be passed through using attention mechanism (Attention mechanism)
Input different conditions vector 340 in each time to solve the problems, such as this, each moment choose with currently to be exported it is most suitable
State vector.Specifically, a can be usedijWeigh jth stage h in Encoder330jWith the i-th stage in Decoder350
Correlation, the state vector of the input in the i-th stage comes from all h in Encoder330 in final Decoder350jTo aij's
Weighted sum.Specifically, aijIt can also be obtained by above-mentioned RNN model learnings and training, thus, it is possible to obtain multiple and different
State vector 340, such as:
C1=h1*a11+h2*a12+h3*a13+...+hn*a1n;
C2=h1*a21+h2*a22+h3*a23+...+hn*a2n;
C3=h1*a31+h2*a32+h3*a33+...+hn*a3n;
...
Cn=h1*an1+h2*an2+h3*an3+...+hn*ann,
Above-mentioned state vector C1To CnCalculating be only schematically for the sake of, the present invention it is not so limited.It is above-mentioned
Attention mechanism can help model to assign different weights to the behavior of list entries, extract important and crucial row
To contribute to model to make more accurate judgement, the iteration optimization of acceleration model.
Further, a kind of user view prediction meanss are additionally provided in this example real-time mode.
Fig. 6 shows the block diagram of user view prediction meanss according to the ... of the embodiment of the present invention.User view prediction meanss 500
Including the first acquisition module 510, training module 520, the second acquisition module 530, prediction module 540.
First acquisition module 510 can be used for acquiring user's history behavior sequence, and the user's history behavior sequence includes
Multiple historical user's behavior items, each historical user's behavior item include historical user's behavior and execute the position of historical user's behavior
Confidence ceases.
Training module 520 can be used for training a user view prediction model according to the user's history behavior sequence.
Second acquisition module 530 can be used for acquiring the real-time behavior sequence of user, and the real-time behavior sequence of user includes
Multiple active user behavior items, each active user behavior item include active user behavior and execute the position of active user behavior
Confidence ceases.
Prediction module 540 can be used for, according to the real-time behavior sequence of the user, passing through the user view prediction model
Predict user behavior.
In the user view prediction meanss of exemplary embodiments of the present invention, it is contemplated that the different location of user behavior
The variation of user view is will produce, makes user behavior when the present invention is in user's history behavior sequence and the real-time behavior sequence of user
The attribute for increasing location information, thereby, it is possible to the transfer and variation in geographical location when dynamically tracking user behavior, via with
The study of family Intention Anticipation model, to predict user behavior according to active user behavior and location information, so as to be based on position
Confidence breath preferably depicts the Demand perference of user.
A specific embodiment according to the present invention, may refer to Fig. 7, and user view prediction meanss 501 are adopted including first
Collect module 510, training module 520, the second acquisition module 530, prediction module 540.
First acquisition module 510 may include behavior item acquisition module 511 and sorting module 512.
Behavior item acquisition module 511 can be used for acquiring multiple historical user's behavior items in a historical time section.
Sorting module 512 can be used for multiple historical user's behavior items in the historical time section by historical user's row
For execution time-sequencing form the user's history behavior sequence.
The present invention acquires multiple history in a historical time section by behavior item acquisition module 511 and sorting module 512
User behavior item, and be trained with this, to realize that the short-term real-time intention of user is portrayed, to improve user's behavior prediction
Precision and accuracy.
Training module 520 may include division module 521, input forms module 522 and output forms module 523.
Division module 521 can be used for for historical user's behavior item being divided into first with first kind historical user's behavior
Class historical user's behavior item and second class historical user's behavior item with second class historical user's behavior.
It, will be with specifically, unrelated historical user's behavior can be converted as first kind historical user's behavior using with commodity
Commodity convert relevant historical user's behavior as second class historical user's behavior.For example, first kind historical user's row
To may include search key, screening category;Second class historical user's behavior may include exposure commodity, click quotient
Product, lower single commodity.
Input forms module 522 and can be used for using the first kind historical user behavior item or the first kind history
The input of family behavior item and the second class historical behavior item as the user view prediction model.
Output forms module 523 and can be used for predicting mould using the second class historical user behavior as the user view
The output of type.
The present invention can form module 522 by division module 521, input and output forms module 523 and determines model
It outputs and inputs, for subsequently carrying out the solution and training of model parameter.
Training module 520 can also include primary vector module 524.
Primary vector module 524 may be used to the user's history behavior sequence vectorization, wherein respectively to described
Historical user's behavior of each historical user's behavior item and the location information carry out vectorization in user's history behavior sequence,
And historical user's behavior by vectorization and location information splice the user's history behavior item for obtaining vectorization.
Prediction module 540 may include secondary vector module 541.
Secondary vector module 541 may be used to the real-time behavior sequence vectorization of the user, wherein respectively to described
The active user behavior of each active user behavior item and the location information carry out vectorization in the real-time behavior sequence of user,
And active user behavior by vectorization and location information splice the real-time behavior item of user for obtaining vectorization.
Fig. 6 and 7 is only to show schematically user view prediction meanss provided by the invention, without prejudice to structure of the present invention
Under the premise of think of, the fractionation of module, increases all within protection scope of the present invention merging.
In an exemplary embodiment of the present invention, a kind of computer readable storage medium is additionally provided, meter is stored thereon with
The circulation of electronic prescription described in any one above-mentioned embodiment may be implemented in calculation machine program, the program when being executed by such as processor
The step of processing method.In some possible embodiments, various aspects of the invention are also implemented as a kind of program production
The form of product comprising program code, when described program product is run on the terminal device, said program code is for making institute
State terminal device execute described in this specification above-mentioned electronic prescription circulation processing method part according to the various examples of the present invention
The step of property embodiment.
Refering to what is shown in Fig. 8, describing the program product for realizing the above method according to the embodiment of the present invention
800, portable compact disc read only memory (CD-ROM) may be used and include program code, and can in terminal device,
Such as it is run on PC.However, the program product of the present invention is without being limited thereto, in this document, readable storage medium storing program for executing can be with
To be any include or the tangible medium of storage program, the program can be commanded execution system, device either device use or
It is in connection.
The arbitrary combination of one or more readable mediums may be used in described program product.Readable medium can be readable letter
Number medium or readable storage medium storing program for executing.Readable storage medium storing program for executing for example can be but be not limited to electricity, magnetic, optical, electromagnetic, infrared ray or
System, device or the device of semiconductor, or the arbitrary above combination.The more specific example of readable storage medium storing program for executing is (non exhaustive
List) include:It is electrical connection, portable disc, hard disk, random access memory (RAM) with one or more conducting wires, read-only
Memory (ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, portable compact disc read only memory
(CD-ROM), light storage device, magnetic memory device or above-mentioned any appropriate combination.
The computer readable storage medium may include the data letter propagated in a base band or as a carrier wave part
Number, wherein carrying readable program code.Diversified forms, including but not limited to electromagnetism may be used in the data-signal of this propagation
Signal, optical signal or above-mentioned any appropriate combination.Readable storage medium storing program for executing can also be any other than readable storage medium storing program for executing
Readable medium, which can send, propagate either transmission for being used by instruction execution system, device or device or
Person's program in connection.The program code for including on readable storage medium storing program for executing can transmit with any suitable medium, packet
Include but be not limited to wireless, wired, optical cable, RF etc. or above-mentioned any appropriate combination.
It can be write with any combination of one or more programming languages for executing the program that operates of the present invention
Code, described program design language include object oriented program language-Java, C++ etc., further include conventional
Procedural programming language-such as " C " language or similar programming language.Program code can be fully in tenant
It is executed on computing device, partly executes in tenant's equipment, executed as an independent software package, partly calculated in tenant
Upper side point is executed or is executed in remote computing device or server completely on a remote computing.It is being related to far
In the situation of journey computing device, remote computing device can pass through the network of any kind, including LAN (LAN) or wide area network
(WAN), it is connected to tenant's computing device, or, it may be connected to external computing device (such as utilize ISP
To be connected by internet).
In an exemplary embodiment of the present invention, a kind of electronic equipment is also provided, which may include processor,
And the memory of the executable instruction for storing the processor.Wherein, the processor is configured to via described in execution
Executable instruction is come the step of executing the circulation processing method of electronic prescription described in any one above-mentioned embodiment.
Person of ordinary skill in the field it is understood that various aspects of the invention can be implemented as system, method or
Program product.Therefore, various aspects of the invention can be embodied in the following forms, i.e.,:It is complete hardware embodiment, complete
The embodiment combined in terms of full Software Implementation (including firmware, microcode etc.) or hardware and software, can unite here
Referred to as circuit, " module " or " system ".
The electronic equipment 600 of this embodiment according to the present invention is described referring to Fig. 9.The electronics that Fig. 9 is shown
Equipment 600 is only an example, should not bring any restrictions to the function and use scope of the embodiment of the present invention.
As shown in figure 9, electronic equipment 600 is showed in the form of universal computing device.The component of electronic equipment 600 can wrap
It includes but is not limited to:At least one processing unit 610, at least one storage unit 620, (including the storage of connection different system component
Unit 620 and processing unit 610) bus 630, display unit 640 etc..
Wherein, the storage unit has program stored therein code, and said program code can be held by the processing unit 610
Row so that the processing unit 610 execute described in this specification above-mentioned electronic prescription circulation processing method part according to this
The step of inventing various illustrative embodiments.For example, the processing unit 610 can execute as shown in Figure 1 to Figure 4 in any figure institute
The step of showing.
The storage unit 620 may include the readable medium of volatile memory cell form, such as random access memory
Unit (RAM) 6201 and/or cache memory unit 6202 can further include read-only memory unit (ROM) 6203.
The storage unit 620 can also include program/practicality work with one group of (at least one) program module 6205
Tool 6204, such program module 6205 include but not limited to:Operating system, one or more application program, other programs
Module and program data may include the realization of network environment in each or certain combination in these examples.
Bus 630 can be to indicate one or more in a few class bus structures, including storage unit bus or storage
Cell controller, peripheral bus, graphics acceleration port, processing unit use the arbitrary bus structures in a variety of bus structures
Local bus.
Electronic equipment 600 can also be with one or more external equipments 700 (such as keyboard, sensing equipment, bluetooth equipment
Deng) communication, can also enable the equipment that tenant interact with the electronic equipment 600 to communicate with one or more, and/or with make
Any equipment that the electronic equipment 600 can be communicated with one or more of the other computing device (such as router, modulation /demodulation
Device etc.) communication.This communication can be carried out by input/output (I/O) interface 650.Also, electronic equipment 600 can be with
By network adapter 660 and one or more network (such as LAN (LAN), wide area network (WAN) and/or public network,
Such as internet) communication.Network adapter 660 can be communicated by bus 630 with other modules of electronic equipment 600.It should
Understand, although not shown in the drawings, other hardware and/or software module can be used in conjunction with electronic equipment 600, including but it is unlimited
In:Microcode, device driver, redundant processing unit, external disk drive array, RAID system, tape drive and number
According to backup storage system etc..
Through the above description of the embodiments, those skilled in the art is it can be readily appreciated that example described herein is implemented
Mode can also be realized by software realization in such a way that software is in conjunction with necessary hardware.Therefore, according to the present invention
The technical solution of embodiment can be expressed in the form of software products, the software product can be stored in one it is non-volatile
Property storage medium (can be CD-ROM, USB flash disk, mobile hard disk etc.) in or network on, including some instructions are so that a calculating
Equipment (can be personal computer, server or network equipment etc.) executes the above-mentioned electronics according to embodiment of the present invention
Prescription circulation processing method.
Compared with prior art, advantage of the invention is that:
In view of the different location of user behavior will produce the variation of user view, the present invention is in user's history behavior sequence
And user behavior is made to increase the attribute of location information when in the real-time behavior sequence of user, thereby, it is possible to dynamically track user's row
For when geographical location transfer and variation, via the study of user view prediction model, with according to active user behavior and position
Information prediction user behavior is set, so as to preferably depict the Demand perference of user based on location information.
Those skilled in the art after considering the specification and implementing the invention disclosed here, will readily occur to its of the present invention
Its embodiment.This application is intended to cover the present invention any variations, uses, or adaptations, these modifications, purposes or
Person's adaptive change follows the general principle of the present invention and includes undocumented common knowledge in the art of the invention
Or conventional techniques.The description and examples are only to be considered as illustrative, and true scope and spirit of the invention are by appended
Claim is pointed out.
Claims (11)
1. a kind of user view prediction technique, which is characterized in that including:
User's history behavior sequence is acquired, the user's history behavior sequence includes multiple historical user's behavior items, each described to go through
History user behavior item includes historical user's behavior and executes the location information of historical user's behavior;
A user view prediction model is trained according to the user's history behavior sequence;
The real-time behavior sequence of user is acquired, the real-time behavior sequence of user includes multiple active user behavior items, each reality
When user behavior item include active user behavior and execute active user behavior location information;And
According to the real-time behavior sequence of the user, user behavior is predicted by the user view prediction model.
2. user view prediction technique as described in claim 1, which is characterized in that instructed according to the user's history behavior sequence
Practicing a user view prediction model includes:
Make the user's history behavior sequence vectorization, wherein respectively to each historical user in the user's history behavior sequence
Historical user's behavior of behavior item and the location information carry out vectorization, and by historical user's behavior of vectorization and to
The location information of quantization splices the user's history behavior item for obtaining vectorization;
The user view prediction model is trained according to the user's history behavior sequence of vectorization.
3. user view prediction technique as claimed in claim 2, which is characterized in that historical user's behavior uses
Word2vev algorithm vectorizations.
4. user view prediction technique as claimed in claim 2, which is characterized in that the position in historical user's behavior item
Information is the character string forms formed using geohash algorithms, the location informations of the character string forms by mapping table to
Quantization.
5. the user view prediction technique as described in any one of claim 1-4, which is characterized in that the user view prediction
Model is Recognition with Recurrent Neural Network RNN, described to include according to the user's history behavior sequence one user view prediction model of training:
Using at least partly described user's history behavior sequence as the input of the Recognition with Recurrent Neural Network RNN;
Using at least partly described user's history behavior sequence as the output of the Recognition with Recurrent Neural Network RNN, wherein the cycle
The output of neural network RNN is converted according to state vector and is obtained, and the state vector is defeated according to the Recognition with Recurrent Neural Network RNN's
Enter conversion to obtain.
6. user view prediction technique as claimed in claim 5, which is characterized in that the Recognition with Recurrent Neural Network RNN includes compiling
Code layer and decoding layer, historical user's behavior item of at least partly described user's history behavior sequence sequentially input the coding layer with
The state vector is obtained according to the first hidden vector of coding layer output, the state vector repeatedly inputs the decoding layer with root
The output of the Recognition with Recurrent Neural Network RNN is obtained according to the second hidden vector that the decoding layer is sequentially output,
Wherein, the coding layer receives the input of j-th of Recognition with Recurrent Neural Network RNN in the jth stage, and according to the coding
First hidden vector h of the layer in the output of -1 stage of jthj-1Calculate the first hidden vector h that the coding layer exports in the jth stagej, j is
The integer of input quantity more than or equal to 1 less than or equal to the Recognition with Recurrent Neural Network RNN;
The decoding layer receives the state vector in the i-th stage, and according to the decoding layer the (i-1)-th stage export second
Hidden vector hi-1' calculate the second hidden vector h that the decoding layer exports in the i-th stagei', and it is based on the described second hidden vector hi' carry
For the output of i-th of Recognition with Recurrent Neural Network RNN, i is defeated less than or equal to the Recognition with Recurrent Neural Network RNN more than or equal to 1
Go out the integer of quantity.
7. user view prediction technique as claimed in claim 6, which is characterized in that the Recognition with Recurrent Neural Network RNN is in difference
To use different state vectors, the state vector that the decoding layer input in the i-th stage be first that coding layer exports stage
Hidden vector hjWith aijWeighted sum, aijIndicate the first hidden vector h in the coding layer jth stagejWith the i-th rank of the decoding layer
The correlation of section.
8. the user view prediction technique as described in any one of claim 1-4,6-7, which is characterized in that according to the user
Historical behavior sequence train a user view prediction model include:
Historical user's behavior item is divided into first kind historical user's behavior item with first kind historical user's behavior and is had
Second class historical user's behavior item of second class historical user's behavior, wherein the first kind historical user behavior is and commodity
Unrelated historical user's behavior is converted, the second class historical user's behavior is to convert relevant historical user's behavior with commodity;
By first kind historical user's behavior item or the first kind historical user behavior item and the second class historical behavior item
Input as the user view prediction model;
Using the second class historical user behavior as the output of the user view prediction model.
9. a kind of user view prediction meanss, which is characterized in that including:
First acquisition module, for acquiring user's history behavior sequence, the user's history behavior sequence includes that multiple history are used
Family behavior item, each historical user's behavior item include historical user's behavior and execute the location information of historical user's behavior;
Training module, for training a user view prediction model according to the user's history behavior sequence;
Second acquisition module, for acquiring the real-time behavior sequence of user, the real-time behavior sequence of user includes multiple real-time use
Family behavior item, each active user behavior item include active user behavior and execute the location information of active user behavior;With
And
Prediction module, for according to the real-time behavior sequence of the user, user's row to be predicted by the user view prediction model
For.
10. a kind of electronic equipment, which is characterized in that the electronic equipment includes:
Processor;
Memory is stored thereon with computer program, is executed when the computer program is run by the processor as right is wanted
Seek 1 to 8 any one of them step.
11. a kind of storage medium, which is characterized in that be stored with computer program, the computer program on the storage medium
Such as claim 1 to 8 any one of them step is executed when being run by processor.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810218834.8A CN108446374B (en) | 2018-03-16 | 2018-03-16 | User's Intention Anticipation method, apparatus, electronic equipment, storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810218834.8A CN108446374B (en) | 2018-03-16 | 2018-03-16 | User's Intention Anticipation method, apparatus, electronic equipment, storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108446374A true CN108446374A (en) | 2018-08-24 |
CN108446374B CN108446374B (en) | 2019-09-20 |
Family
ID=63195596
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810218834.8A Active CN108446374B (en) | 2018-03-16 | 2018-03-16 | User's Intention Anticipation method, apparatus, electronic equipment, storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108446374B (en) |
Cited By (32)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109345260A (en) * | 2018-10-09 | 2019-02-15 | 北京芯盾时代科技有限公司 | A kind of fraud detection model training method and device and fraud detection method and device |
CN109635079A (en) * | 2018-10-25 | 2019-04-16 | 北京中关村科金技术有限公司 | A kind of determination method, apparatus, computer equipment and storage medium that user is intended to |
CN109783736A (en) * | 2019-01-18 | 2019-05-21 | 广东小天才科技有限公司 | Intention presumption method and system |
CN109800359A (en) * | 2018-12-20 | 2019-05-24 | 北京百度网讯科技有限公司 | Information recommendation processing method, device, electronic equipment and readable storage medium storing program for executing |
CN109858970A (en) * | 2019-02-02 | 2019-06-07 | 中国银行股份有限公司 | A kind of user's behavior prediction method, apparatus and storage medium |
CN109901896A (en) * | 2018-12-06 | 2019-06-18 | 华为技术有限公司 | Multi-task processing method in a kind of man-machine interactive system and man-machine interactive system |
CN109947029A (en) * | 2019-04-10 | 2019-06-28 | 广东工业大学 | Control method, device and the equipment of smart home device |
CN110166529A (en) * | 2019-04-16 | 2019-08-23 | 平安普惠企业管理有限公司 | It keeps logging in state method, apparatus, equipment and storage medium |
CN110290582A (en) * | 2019-05-22 | 2019-09-27 | 浙江工业大学 | A kind of base station label trajectory predictions method based on seq2seq frame |
CN110457572A (en) * | 2019-05-23 | 2019-11-15 | 北京邮电大学 | Commodity information recommendation method and electronic equipment based on figure network |
CN110515904A (en) * | 2019-08-13 | 2019-11-29 | 北京达佳互联信息技术有限公司 | Quality prediction model training method, qualitative forecasting method and the device of media file |
CN110874145A (en) * | 2018-08-30 | 2020-03-10 | 北京搜狗科技发展有限公司 | Input method and device and electronic equipment |
CN111199430A (en) * | 2020-01-08 | 2020-05-26 | 中信银行股份有限公司 | Service solution generation method and device and electronic equipment |
CN111258469A (en) * | 2020-01-09 | 2020-06-09 | 支付宝(杭州)信息技术有限公司 | Method and device for processing interactive sequence data |
CN111340112A (en) * | 2020-02-26 | 2020-06-26 | 腾讯科技(深圳)有限公司 | Classification method, classification device and server |
CN111460302A (en) * | 2020-03-31 | 2020-07-28 | 拉扎斯网络科技(上海)有限公司 | Data processing method and device, electronic equipment and computer readable storage medium |
CN111507108A (en) * | 2020-04-17 | 2020-08-07 | 腾讯科技(深圳)有限公司 | Alias generation method and device, electronic equipment and computer readable storage medium |
CN111813640A (en) * | 2020-03-25 | 2020-10-23 | 北京嘀嘀无限科技发展有限公司 | Method, apparatus, electronic device and storage medium for predicting user behavior |
CN111915338A (en) * | 2019-05-08 | 2020-11-10 | 腾讯科技(深圳)有限公司 | Target variable prediction method, device and storage medium |
CN112016316A (en) * | 2020-08-31 | 2020-12-01 | 北京嘀嘀无限科技发展有限公司 | Identification method and system |
CN112115258A (en) * | 2019-06-20 | 2020-12-22 | 腾讯科技(深圳)有限公司 | User credit evaluation method, device, server and storage medium |
CN112150177A (en) * | 2019-06-27 | 2020-12-29 | 百度在线网络技术(北京)有限公司 | Intention prediction method and device |
CN112182434A (en) * | 2020-09-28 | 2021-01-05 | 北京红山信息科技研究院有限公司 | User portrait generation method, system, equipment and storage medium |
CN112241327A (en) * | 2019-07-17 | 2021-01-19 | 腾讯科技(深圳)有限公司 | Shared information processing method and device, storage medium and electronic equipment |
CN112700277A (en) * | 2020-12-30 | 2021-04-23 | 北京爱笔科技有限公司 | User behavior data processing method and multi-behavior sequence conversion model training method |
CN112801719A (en) * | 2021-03-01 | 2021-05-14 | 深圳市欢太科技有限公司 | User behavior prediction method, user behavior prediction device, storage medium, and apparatus |
CN113015225A (en) * | 2021-02-19 | 2021-06-22 | 清华大学 | Wireless network resource arranging method and device for machine-oriented communication |
CN113298321A (en) * | 2021-06-22 | 2021-08-24 | 深圳市查策网络信息技术有限公司 | User intention prediction method based on multi-data fusion |
CN113781079A (en) * | 2020-10-13 | 2021-12-10 | 北京沃东天骏信息技术有限公司 | Method and apparatus for training a model |
CN114386688A (en) * | 2022-01-07 | 2022-04-22 | 深圳市聚商鼎力网络技术有限公司 | User intention prediction method and system based on multi-data fusion |
WO2022222854A1 (en) * | 2021-04-18 | 2022-10-27 | 华为技术有限公司 | Data processing method and related device |
CN115757900A (en) * | 2022-12-20 | 2023-03-07 | 邢台达喆网络科技有限公司 | User demand analysis method and system applying artificial intelligence model |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102117325A (en) * | 2011-02-24 | 2011-07-06 | 清华大学 | Method for predicting dynamic social network user behaviors |
CN102346899A (en) * | 2011-10-08 | 2012-02-08 | 亿赞普(北京)科技有限公司 | Method and device for predicting advertisement click rate based on user behaviors |
CN105868281A (en) * | 2016-03-23 | 2016-08-17 | 西安电子科技大学 | Location-aware recommendation system based on non-dominated sorting multi-target method |
US20170213006A1 (en) * | 2016-01-27 | 2017-07-27 | The Live Network Inc. | Intelligent mobile homework adherence and feedback application for telehealth |
CN107578294A (en) * | 2017-09-28 | 2018-01-12 | 北京小度信息科技有限公司 | User's behavior prediction method, apparatus and electronic equipment |
-
2018
- 2018-03-16 CN CN201810218834.8A patent/CN108446374B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102117325A (en) * | 2011-02-24 | 2011-07-06 | 清华大学 | Method for predicting dynamic social network user behaviors |
CN102346899A (en) * | 2011-10-08 | 2012-02-08 | 亿赞普(北京)科技有限公司 | Method and device for predicting advertisement click rate based on user behaviors |
US20170213006A1 (en) * | 2016-01-27 | 2017-07-27 | The Live Network Inc. | Intelligent mobile homework adherence and feedback application for telehealth |
CN105868281A (en) * | 2016-03-23 | 2016-08-17 | 西安电子科技大学 | Location-aware recommendation system based on non-dominated sorting multi-target method |
CN107578294A (en) * | 2017-09-28 | 2018-01-12 | 北京小度信息科技有限公司 | User's behavior prediction method, apparatus and electronic equipment |
Cited By (48)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110874145A (en) * | 2018-08-30 | 2020-03-10 | 北京搜狗科技发展有限公司 | Input method and device and electronic equipment |
CN109345260B (en) * | 2018-10-09 | 2021-11-30 | 北京芯盾时代科技有限公司 | Method for detecting abnormal operation behavior |
CN109345260A (en) * | 2018-10-09 | 2019-02-15 | 北京芯盾时代科技有限公司 | A kind of fraud detection model training method and device and fraud detection method and device |
CN109635079A (en) * | 2018-10-25 | 2019-04-16 | 北京中关村科金技术有限公司 | A kind of determination method, apparatus, computer equipment and storage medium that user is intended to |
US20210166693A1 (en) * | 2018-12-06 | 2021-06-03 | Huawei Technologies Co., Ltd. | Man- machine interaction system and multi-task processing method in the man-machine interaction system |
CN109901896A (en) * | 2018-12-06 | 2019-06-18 | 华为技术有限公司 | Multi-task processing method in a kind of man-machine interactive system and man-machine interactive system |
US12039976B2 (en) | 2018-12-06 | 2024-07-16 | Huawei Technologies Co., Ltd. | Man- machine interaction system and multi-task processing method in the man-machine interaction system |
CN109800359B (en) * | 2018-12-20 | 2021-08-17 | 北京百度网讯科技有限公司 | Information recommendation processing method and device, electronic equipment and readable storage medium |
CN109800359A (en) * | 2018-12-20 | 2019-05-24 | 北京百度网讯科技有限公司 | Information recommendation processing method, device, electronic equipment and readable storage medium storing program for executing |
CN109783736B (en) * | 2019-01-18 | 2022-03-08 | 广东小天才科技有限公司 | Intention presumption method and system |
CN109783736A (en) * | 2019-01-18 | 2019-05-21 | 广东小天才科技有限公司 | Intention presumption method and system |
CN109858970B (en) * | 2019-02-02 | 2021-07-02 | 中国银行股份有限公司 | User behavior prediction method, device and storage medium |
CN109858970A (en) * | 2019-02-02 | 2019-06-07 | 中国银行股份有限公司 | A kind of user's behavior prediction method, apparatus and storage medium |
CN109947029A (en) * | 2019-04-10 | 2019-06-28 | 广东工业大学 | Control method, device and the equipment of smart home device |
CN110166529A (en) * | 2019-04-16 | 2019-08-23 | 平安普惠企业管理有限公司 | It keeps logging in state method, apparatus, equipment and storage medium |
CN111915338B (en) * | 2019-05-08 | 2024-05-14 | 腾讯科技(深圳)有限公司 | Target variable prediction method, device and storage medium |
CN111915338A (en) * | 2019-05-08 | 2020-11-10 | 腾讯科技(深圳)有限公司 | Target variable prediction method, device and storage medium |
CN110290582B (en) * | 2019-05-22 | 2020-10-30 | 浙江工业大学 | Base station label track prediction method based on seq2seq frame |
CN110290582A (en) * | 2019-05-22 | 2019-09-27 | 浙江工业大学 | A kind of base station label trajectory predictions method based on seq2seq frame |
CN110457572B (en) * | 2019-05-23 | 2022-05-24 | 北京邮电大学 | Commodity information recommendation method based on graph network and electronic equipment |
CN110457572A (en) * | 2019-05-23 | 2019-11-15 | 北京邮电大学 | Commodity information recommendation method and electronic equipment based on figure network |
CN112115258B (en) * | 2019-06-20 | 2023-09-26 | 腾讯科技(深圳)有限公司 | Credit evaluation method and device for user, server and storage medium |
CN112115258A (en) * | 2019-06-20 | 2020-12-22 | 腾讯科技(深圳)有限公司 | User credit evaluation method, device, server and storage medium |
CN112150177A (en) * | 2019-06-27 | 2020-12-29 | 百度在线网络技术(北京)有限公司 | Intention prediction method and device |
CN112241327A (en) * | 2019-07-17 | 2021-01-19 | 腾讯科技(深圳)有限公司 | Shared information processing method and device, storage medium and electronic equipment |
CN110515904A (en) * | 2019-08-13 | 2019-11-29 | 北京达佳互联信息技术有限公司 | Quality prediction model training method, qualitative forecasting method and the device of media file |
CN111199430A (en) * | 2020-01-08 | 2020-05-26 | 中信银行股份有限公司 | Service solution generation method and device and electronic equipment |
CN111258469A (en) * | 2020-01-09 | 2020-06-09 | 支付宝(杭州)信息技术有限公司 | Method and device for processing interactive sequence data |
CN111340112B (en) * | 2020-02-26 | 2023-09-26 | 腾讯科技(深圳)有限公司 | Classification method, classification device and classification server |
CN111340112A (en) * | 2020-02-26 | 2020-06-26 | 腾讯科技(深圳)有限公司 | Classification method, classification device and server |
CN111813640A (en) * | 2020-03-25 | 2020-10-23 | 北京嘀嘀无限科技发展有限公司 | Method, apparatus, electronic device and storage medium for predicting user behavior |
CN111460302B (en) * | 2020-03-31 | 2023-08-08 | 拉扎斯网络科技(上海)有限公司 | Data processing method, device, electronic equipment and computer readable storage medium |
CN111460302A (en) * | 2020-03-31 | 2020-07-28 | 拉扎斯网络科技(上海)有限公司 | Data processing method and device, electronic equipment and computer readable storage medium |
CN111507108B (en) * | 2020-04-17 | 2021-03-19 | 腾讯科技(深圳)有限公司 | Alias generation method and device, electronic equipment and computer readable storage medium |
CN111507108A (en) * | 2020-04-17 | 2020-08-07 | 腾讯科技(深圳)有限公司 | Alias generation method and device, electronic equipment and computer readable storage medium |
CN112016316A (en) * | 2020-08-31 | 2020-12-01 | 北京嘀嘀无限科技发展有限公司 | Identification method and system |
CN112182434A (en) * | 2020-09-28 | 2021-01-05 | 北京红山信息科技研究院有限公司 | User portrait generation method, system, equipment and storage medium |
CN113781079A (en) * | 2020-10-13 | 2021-12-10 | 北京沃东天骏信息技术有限公司 | Method and apparatus for training a model |
CN112700277B (en) * | 2020-12-30 | 2024-05-28 | 北京爱笔科技有限公司 | Processing method of user behavior data and multi-behavior sequence conversion model training method |
CN112700277A (en) * | 2020-12-30 | 2021-04-23 | 北京爱笔科技有限公司 | User behavior data processing method and multi-behavior sequence conversion model training method |
CN113015225A (en) * | 2021-02-19 | 2021-06-22 | 清华大学 | Wireless network resource arranging method and device for machine-oriented communication |
CN112801719A (en) * | 2021-03-01 | 2021-05-14 | 深圳市欢太科技有限公司 | User behavior prediction method, user behavior prediction device, storage medium, and apparatus |
WO2022222854A1 (en) * | 2021-04-18 | 2022-10-27 | 华为技术有限公司 | Data processing method and related device |
CN113298321A (en) * | 2021-06-22 | 2021-08-24 | 深圳市查策网络信息技术有限公司 | User intention prediction method based on multi-data fusion |
CN113298321B (en) * | 2021-06-22 | 2022-03-11 | 深圳市查策网络信息技术有限公司 | User intention prediction method based on multi-data fusion |
CN114386688A (en) * | 2022-01-07 | 2022-04-22 | 深圳市聚商鼎力网络技术有限公司 | User intention prediction method and system based on multi-data fusion |
CN115757900B (en) * | 2022-12-20 | 2023-08-01 | 创贸科技(深圳)集团有限公司 | User demand analysis method and system applying artificial intelligent model |
CN115757900A (en) * | 2022-12-20 | 2023-03-07 | 邢台达喆网络科技有限公司 | User demand analysis method and system applying artificial intelligence model |
Also Published As
Publication number | Publication date |
---|---|
CN108446374B (en) | 2019-09-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108446374B (en) | User's Intention Anticipation method, apparatus, electronic equipment, storage medium | |
JP7206288B2 (en) | Music recommendation method, apparatus, computing equipment and medium | |
CN111339415B (en) | Click rate prediction method and device based on multi-interactive attention network | |
CN111046286B (en) | Object recommendation method and device and computer storage medium | |
CN110162690A (en) | Determine user to the method and apparatus of the interest-degree of article, equipment and storage medium | |
CN109992710A (en) | Clicking rate predictor method, system, medium and calculating equipment | |
CN109885756B (en) | CNN and RNN-based serialization recommendation method | |
CN115917535A (en) | Recommendation model training method, recommendation device and computer readable medium | |
JP2016532962A (en) | Recommended results display method and equipment | |
CN109783539A (en) | Usage mining and its model building method, device and computer equipment | |
CN111950593A (en) | Method and device for recommending model training | |
CN113254711B (en) | Interactive image display method and device, computer equipment and storage medium | |
CN113781149B (en) | Information recommendation method and device, computer readable storage medium and electronic equipment | |
CN111046170A (en) | Method and apparatus for outputting information | |
CN114663197A (en) | Commodity recommendation method and device, equipment, medium and product thereof | |
CN109034853A (en) | Similar users method, apparatus, medium and electronic equipment are found based on seed user | |
CN117216535A (en) | Training method, device, equipment and medium for recommended text generation model | |
CN114706987A (en) | Text category prediction method, device, equipment, storage medium and program product | |
CN112989174A (en) | Information recommendation method and device, medium and equipment | |
CN112269943B (en) | Information recommendation system and method | |
CN114817692A (en) | Method, device and equipment for determining recommended object and computer storage medium | |
CN116910357A (en) | Data processing method and related device | |
CN111768218B (en) | Method and device for processing user interaction information | |
CN114429384B (en) | Intelligent product recommendation method and system based on e-commerce platform | |
CN116186541A (en) | Training method and device for recommendation model |
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 |