CN105447730A - Target user orientation method and device - Google Patents
Target user orientation method and device Download PDFInfo
- Publication number
- CN105447730A CN105447730A CN201510996108.5A CN201510996108A CN105447730A CN 105447730 A CN105447730 A CN 105447730A CN 201510996108 A CN201510996108 A CN 201510996108A CN 105447730 A CN105447730 A CN 105447730A
- Authority
- CN
- China
- Prior art keywords
- user
- users
- seed
- sample
- training
- 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
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000012549 training Methods 0.000 claims abstract description 111
- 238000012360 testing method Methods 0.000 claims description 21
- 239000000284 extract Substances 0.000 claims description 16
- 238000011156 evaluation Methods 0.000 claims description 13
- 230000007115 recruitment Effects 0.000 claims description 13
- 230000006399 behavior Effects 0.000 description 17
- 238000006243 chemical reaction Methods 0.000 description 10
- 230000000694 effects Effects 0.000 description 8
- 230000006870 function Effects 0.000 description 7
- 238000004891 communication Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000011524 similarity measure Methods 0.000 description 3
- 235000010627 Phaseolus vulgaris Nutrition 0.000 description 2
- 244000046052 Phaseolus vulgaris Species 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 239000012141 concentrate Substances 0.000 description 2
- 230000008878 coupling Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000010365 information processing Effects 0.000 description 2
- 230000004807 localization Effects 0.000 description 2
- 230000033001 locomotion Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 241001632422 Radiola linoides Species 0.000 description 1
- 230000019552 anatomical structure morphogenesis Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000004615 ingredient Substances 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
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
- G06Q30/0241—Advertisements
- G06Q30/0251—Targeted advertisements
- G06Q30/0269—Targeted advertisements based on user profile or attribute
- G06Q30/0271—Personalized advertisement
Landscapes
- Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Engineering & Computer Science (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the invention discloses a target user orientation method and device. The method comprises: generating a virtual user basing on the user characteristic of a seed user; taking the seed user and the virtual user as positive sample users; determining negative sample users; respectively extracting the user characteristics of the positive sample users and the negative sample users; training basing on the user characteristics of the positive sample users and the negative sample users, determining orientation parameters; and orientating the target user satisfying the predetermined similarity condition with the seed user basing on the orientation parameters.
Description
Technical field
The present invention relates to areas of information technology, particularly relate to a kind of targeted customer's orientation method and device.
Background technology
Along with the development of infotech, determine that in information pushing field the targeted customer of information pushing is the problem always endeavouring in prior art to solve to promote the efficiency of information pushing.Described information pushing comprises advertisement pushing, to video, audio frequency and graph text information etc. that user recommends.This time, in order to reduce the dislike of user to information pushing, promotes reception and registration and the conversion efficiency of user's user satisfaction and pushed information, usually needs selection to the interested user of pushed information, may just occur localizing objects user.Provide the method for following several localizing objects user in the prior art:
The first: user will be transformed in a period of time as seed user, as positive sample of users; Using unconverted user as negative sample user; Then to user query behavior, the user characteristicses such as web page browsing behavior and social data extract, and the user behavior according to extracting obtains transformation model; Utilize this transformation model to estimate all users, determine whether the audient into this advertisement.
The second: utilize the user that conversion behavior occurred as positive sample of users, other users obtain a primary election model as negative sample user; This feature that user location crosses the webpage of advertisement mainly paid close attention to by this model.Selected model also can be utilized to carry out localizing objects user, in selected model, not only can pay close attention to user whether accessed corresponding website, also can pay close attention to the more attribute tags of user.Here attribute tags can comprise tag along sort, comprises the feature etc. associated with advertiser website that user pays close attention to.Localizing objects user is carried out in conjunction with primary election model and selected model.
The third: first utilize the data pick-up user characteristicses such as user registers, behavior, social activity, and set up inverted index, index entry is feature, and index value is user, regular update; Then, after advertiser's given seed user number bag, on the user characteristics basis extracted, the higher character subset of correlativity is filtered out based on feature selection approachs such as mutual informations; The character subset structure retrieval and inquisition similar users finally using correlativity high, the process of similar search engine. retrieves similar document.Advertiser does portrait analysis according to the similar users of expansion, determines whether use this function.Except being provided in line retrieval expansion similar users, additionally provide off-line mining model, same, in given seed user number bag high correlation feature base, local sensitive hash cluster being done to user, excavating similar users by calculating Hamming distances.
Although said method achieves the location of targeted customer, the degree of accuracy of targeted customer location is still very low, exists the identification of non-targeted user error in order to targeted customer or omit real targeted customer after adopting said method localizing objects user.
Summary of the invention
In view of this, the embodiment of the present invention is expected to provide a kind of targeted customer's localization method and device, solves the accurate not problem in targeted customer location at least partly.
For achieving the above object, technical scheme of the present invention is achieved in that
Embodiment of the present invention first aspect provides a kind of targeted customer's orientation method, and described method comprises:
Based on the user characteristics of seed user, generating virtual user;
Using described seed user and described Virtual User jointly as positive sample of users;
Determine negative sample user;
Extract the user characteristics of described positive sample of users and described negative sample user respectively;
User characteristics based on described positive sample of users and described negative sample user is trained, and determines orientation parameter;
Based on described orientation parameter, orient the targeted customer meeting predetermined similarity condition with described seed user.
Based on such scheme, the described user characteristics based on seed user, generating virtual user, comprising:
Determine the primary importance of described seed user at feature space;
Based at least two described primary importance determination second places;
The user characteristics of described Virtual User is determined and generating virtual user based on the described second place.
Based on such scheme, the described user characteristics based on seed user, generating virtual user, comprising:
Extract the numerical value user characteristics in the user characteristics of described seed user;
Based on described numerical value user characteristics, determine the range of choices of numerical value user characteristics;
The numerical value user characteristics of a numerical value as described Virtual User is selected from described range of choices;
Extract the nonumeric user characteristics in the user characteristics of described seed user, and give probable value to described nonumeric user characteristics;
The nonumeric user characteristics of described Virtual User is determined based on described probable value.
Based on such scheme, describedly determine negative sample user, comprising:
Based on the user characteristics of described seed user, calculate the similarity of alternative sample of users and described seed user;
Based on described similarity, determine the described alternative sample of users meeting negative sample user condition, as the negative sample user in described sample of users.
Based on such scheme, the described user characteristics based on described positive sample of users and described negative sample user is trained, and determines orientation parameter, comprising:
Utilize the user characteristics of described positive sample of users and described negative sample user to carry out model training, determine the disaggregated model of select target user.
Based on such scheme, the described user characteristics of described sample of users that utilizes carries out model training, determines the training pattern of select target user, comprising:
Described positive sample of users and negative sample user are divided into training set, checking collection and test set;
Utilize described training set, adopt different training algorithms to carry out model training;
Described checking is utilized to collect checking to described model training the need of continuation;
After the described model training of stopping, described test set is utilized to carry out recruitment evaluation to the alternative model that each described training algorithm obtains;
Based on described recruitment evaluation, select a described alternative model as described disaggregated model.
Embodiment of the present invention second aspect provides a kind of targeted customer's orienting device, and described device comprises:
Generation unit, for the user characteristics based on seed user, generating virtual user;
First determining unit, for using described seed user and described Virtual User jointly as positive sample of users;
Second determining unit, for determining negative sample user;
Training unit, for extracting the user characteristics of described positive sample of users and described negative sample user respectively;
Positioning unit, for based on described orientation parameter, orients the targeted customer meeting predetermined similarity condition with described seed user.
Based on such scheme, described generation unit, for determining the primary importance of described seed user at feature space; Based at least two described primary importance determination second places; The user characteristics of described Virtual User is determined and generating virtual user based on the described second place.
Based on such scheme, described generation unit, specifically for extracting the numerical value user characteristics in the user characteristics of described seed user; Based on described numerical value user characteristics, determine the range of choices of numerical value user characteristics; The numerical value user characteristics of a numerical value as described Virtual User is selected from described range of choices; Extract the nonumeric user characteristics in the user characteristics of described seed user, and give probable value to described nonumeric user characteristics; The nonumeric user characteristics of described Virtual User is determined based on described probable value.
Based on such scheme, described second determining unit, specifically for the user characteristics based on described seed user, calculates the similarity of alternative sample of users and described seed user; Based on described similarity, determine the described alternative sample of users meeting negative sample user condition, as the negative sample user in described sample of users.
Based on such scheme, described training unit, specifically for utilizing the user characteristics of described positive sample of users and described negative sample user to carry out model training, determines the disaggregated model of select target user.
Based on such scheme, described training unit, specifically for being divided into training set, checking collection and test set by described positive sample of users and negative sample user; Utilize described training set, adopt different training algorithms to carry out model training; Described checking is utilized to collect checking to described model training the need of continuation; After the described model training of stopping, described test set is utilized to carry out recruitment evaluation to the alternative model that each described training algorithm obtains; Based on described recruitment evaluation, select a described alternative model as described disaggregated model.
Targeted customer's localization method and device described in the embodiment of the present invention, first by the user characteristics generating virtual user based on seed user, using seed user and Virtual User jointly as positive sample of users, the quantity of the positive sample of users of obvious increase, obviously can alleviate because positive sample of users quantity is few that negative sample quantity is many, the training data imbalance caused is topic, and then factor data imbalance can be alleviated cause training the discriminating of the orientation parameter obtained can be by force enough, and the phenomenon that targeted customer's setting accuracy is low.
Accompanying drawing explanation
The schematic flow sheet of the first targeted customer's orientation method that Fig. 1 provides for the embodiment of the present invention;
The schematic flow sheet of the positive sample of users of determination that Fig. 2 provides for the embodiment of the present invention;
The generation schematic diagram of the first Virtual User that Fig. 3 provides for the embodiment of the present invention;
The generation schematic diagram of the second Virtual User that Fig. 4 provides for the embodiment of the present invention;
Fig. 5 A provide for the embodiment of the present invention the first determine the schematic flow sheet of negative sample user;
The schematic flow sheet of the second determination negative sample user that Fig. 5 B provides for the embodiment of the present invention;
The determination method flow schematic diagram of the orientation parameter that Fig. 6 provides for the embodiment of the present invention;
The schematic flow sheet of the second targeted customer orientation method that Fig. 7 provides for the embodiment of the present invention;
The structural representation of a kind of targeted customer's orienting device that Fig. 8 provides for the embodiment of the present invention;
The structural representation of another kind of targeted customer's orienting device that Fig. 9 provides for the embodiment of the present invention.
Embodiment
Below in conjunction with Figure of description and specific embodiment technical scheme of the present invention done and further elaborate.
Embodiment one:
As shown in Figure 1, the present embodiment provides a kind of targeted customer's orientation method, and described method comprises:
Step S110: based on the user characteristics of seed user, generating virtual user;
Step S120: using described seed user and described Virtual User jointly as positive sample of users;
Step S130: determine negative sample user;
Step S140: the user characteristics extracting described positive sample of users and described negative sample user respectively;
Step S150: based on described orientation parameter, orients the targeted customer meeting predetermined similarity condition with described seed user.
Targeted customer's orientation method described in the present embodiment, under can being used in the application scenarios of targeted customer's orientation of targeted customer's orientation of advertisement or the recommendation of the multimedia messages such as video, audio frequency.
Described seed user is the user performing goal activities, such as can comprise at advertisement field and click advertisement, or have purchased the commodity of advertisement or downloaded the application of advertisement, namely perform the user of the conversion behavior that advertisement is expected, or evaluate the user that very large probability performs described goal activities.
Seed user described in the present embodiment can be to extract from peripheral hardware and receives, also can for voluntarily by user that data processing is determined.Such as, needing releasing advertisements A, if advertisement A issues first, then by obtaining the history ad data of the advertisement B very similar to advertisement A, the seed user of user as advertisement A advertisement A being performed to corresponding conversion behavior can be extracted.Certainly, advertisement A has tested and has thrown in a period of time before, and effect data is thrown in the test can extracting advertisement A, extracts the user of the conversion behavior clicking advertisement A or perform advertisement A expectation as described seed user.
Can based on the user characteristics of seed user in step S110 described in the present embodiment, generating virtual user, in the step s 120 using seed user and Virtual User jointly as positive sample of users, obviously increase the number of positive sample of users.Such as, current seed user is M, generates N number of Virtual User in step s 110 according to the user characteristics of seed user, and like this, the sum of positive sample of users is just M+N, is obviously the number M's being greater than seed user.Therefore relative to prior art only using seed user as positive sample of users, obviously the quantity of card sample of users is extended, thus add the data of positive sample of users, thus can reduce because positive sample of users data are few, the discriminating of the orientation parameter obtained caused can be by force enough, the coarse phenomenon in location of the targeted customer caused.
The Virtual User generated based on seed user is in step s 110 all the user having very large similarity with described seed user, and such as, the similarity of described Virtual User and described seed user is greater than similar threshold value.Such as, the age of described Virtual User in this time can all between described 20 to 25 between 20 to 25 years old the age of seed user.Like this, the similarity of obvious Virtual User and seed user is higher.
Usually have viewed same advertisement or all perform advertisement conversion behavior seed user between have a lot of co-user feature, and the Virtual User that the present embodiment generates can be the user very similar to described seed user, this Virtual User corresponds to the user of necessary being, but due to the user that the statistics of database is omitted, also can be pure virtual user, not this user; But because the similarity with seed user is very large, the probability of this user conversion behavior of watching same advertisement or performing advertisement the same as seed user is very large, therefore can, as positive sample of users, be the data source of the orientation parameter of training directional aim user.
As shown in Figure 2, the seed user that solid line represents, by user development, expands out Virtual User represented by dashed line.User's number in the positive sample of users set that seed user and Virtual User form jointly is the number of seed user and the number sum of Virtual User, is obviously greater than the number of seed user.
Any one method can be adopted in step s 130, which to determine negative sample user, such as, any one user beyond described seed user can be selected as described negative sample user.It should be noted that there is no certain precedence relationship between described step S110 and described step S130 in the present embodiment.Described step S110 can perform prior to described step S130, also can perform after described step S130.
In the present embodiment, because positive sample of users is here the user very similar to described seed user, be the user having very high probability to be called described targeted customer.Such as described in the present embodiment, targeted customer's orientation method is applied to advertisement field, described targeted customer is the propelling movement user of advertisement, then described positive sample of users is watch, click and/or buy the commodity of described advertising or the user of service to a great extent, namely has the user that greater probability performs the conversion behavior that described advertisement is expected.Here conversion behavior can comprise clicks the page that lands that advertisement enters advertisement; Buying commodity or service, the application of downloads ad or the activity of participation advertising of advertisement, proposing feedback opinion etc. as given the brand of advertising.
After determining positive sample of users and negative sample user, to aligning sample of users respectively and negative sample user trains, described recursive model will be obtained in step S140.Here the training how carrying out the user characteristics of positive sample of users and negative sample user can adopt the various training tool such as neural network, learning machine to train, and finally obtains the orientation parameter that user can be divided into targeted customer and non-targeted user.
When specific implementation, such as, directed with the targeted customer carrying out advertisement in the network platforms such as social networks, it is one or more that described user characteristics can comprise in user property feature, user behavior feature, user social contact feature.Described user property feature can comprise user's age, user's sex, the education degree of user, the user profile such as education background, the occupation of user, the hobby of user of user.Some user operations that the application (Application, App) that the public number that described user behavior feature can comprise content that user issues in social platform, user pays close attention to, user download, user utilize App to perform.Here the content issued can comprise the original content that user issues, and also can comprise the content that user forwards.Here some user operations that user utilizes App to perform, can such as, the video that user utilizes video App to watch.Analyze the information such as the performer that these videos comprise, can determine that whether user is to the bean vermicelli being some performers, if user is the bean vermicelli of this performer, then recommend to this user the advertisement comprising this performer, just may obtain higher transition probability.Certainly, user behavior feature here can be divided into user's long-term action to be accustomed to according to the time again and user's acts and efforts for expediency deflection etc.User social contact feature can comprise the information such as the social networks chain of user, the social frequency of user and other users, social platform that user is usual.These features can both reflect demand and the interested content of user of user, can as the parameter of directional aim user.
Finally, determine in step S150 be meet default similarity condition with seed user user as targeted customer, like this, targeted customer is because of the similarity with seed user, by very high for the probability performed with seed user similar active or same campaign, can as the targeted customer performing a certain item goal activities.Such as to targeted customer's advertisement etc.
In a word, present embodiments provide a kind of targeted customer's orientation method, be not limited to the quantity of the seed user of acquisition, but the expansion of sample of users automatically can be carried out according to the user characteristics of seed user, obtain more multisample user or more can be used for the sample of users of accurate localizing objects user, improve the setting accuracy of targeted customer.
Embodiment two:
As shown in Figure 1, the present embodiment provides a kind of targeted customer's orientation method, and described method comprises: step S110: based on the user characteristics of seed user, generating virtual user; Here seed user can be the user of advertiser by providing, also can for the user determined according to the history ad data of advertisement; Step S120: using described seed user and described Virtual User jointly as positive sample of users; Like this, the expansion of positive sample of users is just achieved by the generation of Virtual User; Step S130: determine negative sample user; Such as can user beyond Stochastic choice seed user as described negative sample user, certain customers also can be selected from negative sample user as described negative sample user; Step S140: the user characteristics extracting described positive sample of users and described negative sample user respectively; Step S150: based on described orientation parameter, orients the targeted customer meeting predetermined similarity condition with described seed user.
In the present embodiment, described step S110 can comprise:
Determine the primary importance of M described seed user at feature space; Based at least two described primary importance determination second places; The user characteristics of described Virtual User is determined and generating virtual user based on the described second place.
Feature space described in the present embodiment can be the vector space of a multidimensional, in this vector space, seed user is because its similarity, can at least assemble in a dimension or multiple dimension, concentration range can be determined in the dimension of satisfied gathering in the present embodiment, in the white space of concentration range, build described Virtual User.Meet focus on dimension in, can build at random Virtual User user characteristics or, the user characteristics still based on seed user builds.
In the present embodiment, two described primary importances can be the extreme position of concentration range.
As shown in Figure 3, in the vector space of bidimensional, according to the user characteristics of seed user, determine the feature of seed user.Here primary importance and the second place can the coordinate of vector space represent.As can be seen from Figure 4, seed user is assembled near the A of position, and it may be seed user that these users in the white space of position A annex have very large, but due to may when determining seed user, the problems such as data volume is large not, miss these users.Step S110 is utilized to construct Virtual User in the present embodiment.These Virtual User and seed user are formed positive sample of users jointly, to expand the quantity of positive sample of users.Utilize more positive sample of users to provide user characteristics to determine targeted customer, more parameter or parameter more accurately can be provided for targeted customer location, thus the positioning accurate accuracy of targeted customer can be promoted.
Embodiment three:
As shown in Figure 1, the present embodiment provides a kind of targeted customer's orientation method, and described method comprises: step S110: based on the user characteristics of seed user, generating virtual user; Here seed user can be the user of advertiser by providing, also can for the user determined according to the history ad data of advertisement; Step S120: using described seed user and described Virtual User jointly as positive sample of users; Like this, the expansion of positive sample of users is just achieved by the generation of Virtual User; Step S130: determine negative sample user; Such as can user beyond Stochastic choice seed user as described negative sample user, certain customers also can be selected from negative sample user as described negative sample user; Step S140: the user characteristics extracting described positive sample of users and described negative sample user respectively; Step S150: based on described orientation parameter, orients the targeted customer meeting predetermined similarity condition with described seed user.
In the present embodiment, described step S110 can comprise: extract the numerical value user characteristics in the user characteristics of described seed user; Based on described numerical value user characteristics, determine the range of choices of numerical value user characteristics; The numerical value user characteristics of a numerical value as described Virtual User is selected from described range of choices; Extract the nonumeric user characteristics in the user characteristics of described seed user, and give probable value to described nonumeric user characteristics; The nonumeric user characteristics of described Virtual User is determined based on described probable value.
Described numerical value user characteristics is for can quantize user characteristics in the present embodiment, and the age of such as user, user watch the information such as the frequency of a certain class video.Like this, the numerical value user characteristics of seed user will form data interval.Such as, by Information Statistics, find that the age of seed user is all between 15 to 23 years old.Here 15 to 23 are described data interval, are also aforesaid range of choices.Like this when generating virtual user, can an age be selected as the age characteristics of described Virtual User between 15 to 23 years old.Specifically how to select, can determine according to default choice function.Such as described default choice function can be random function.
What described nonumeric user characteristics was corresponding is not quantifiable user characteristics.Such as, each seed user has 3 hobbies, and this hobby sorts corresponding probability successively.Such as, the interest of seed user A is for running, drawing a picture, listen to the music, and probability corresponding is respectively a1, a2, a3; The interest of seed user B is for singing, running, draw a picture, and probability corresponding is respectively b1, b2, b3; Based on seed user A and seed user B, the hobby building Virtual User C can comprise: run, draw a picture, listen to the music, sing; And probability corresponding is respectively followed successively by (a1+b2)/2, (a2+b2)/2, a3/2, a1/2; Again according to probability sorting, the forward one or more hobbies of selected and sorted are the hobby of described Virtual User C.Such as, selected and sorted forward 3 or 2 hobbies are as the hobby of described Virtual User C.That if extract seed user p hobby, then the hobby number of generating virtual user C is not more than p in specific implementation.Described p be less than 1 integer.Such as, the geographic position that the frequency of occurrences of user is higher, can utilize the mode of above-mentioned hobby to determine the geographic position that the motion frequency of Virtual User is higher equally.
Complete the numerical value user characteristics of a Virtual User and the generation of nonumeric user characteristics, just complete the determination of the user characteristics of a Virtual User, then user ID is just complete defines a Virtual User enough similar to seed user for this Virtual User is formed.
In concrete implementation procedure, in step S110 during generating virtual user, can based on the user characteristics of whole seed user, also can only based on the seed user of more than 2 or 2 of Some seeds user.Such as, Virtual User A generates based on the user characteristics of seed user A1, seed user A2 and seed user A3, then this Virtual User A and seed user A1, seed user A2 and seed user A3 are closely similar.But seed user is except seed user A1, seed user A2 and seed user A3, the seed user such as seed user B and seed user C also can be comprised.
In concrete implementation procedure, be not limited to the generation of the Virtual User that embodiment two and embodiment three provide, the present embodiment also provides a kind of geometry to build method, comprising:
With the first child user for origin, build between the first child user and other seed user; Line; Here the first child user can be any one user in seed user;
Each vector gets cut off, utilizes the vector relations between this cut off and the first child user, build Virtual User.The user characteristics of Virtual User here, depends on and distance between this cut off and two seed user.In order to carry out the expansion of seed user as much as possible, at least one half seed user in all seed user can be carried out the structure of Virtual User as the first child user described by poll successively.
As shown in Figure 4, between the first child user and the second child user, define line segment A, line segment A has got a cut off B at random, determine the similarity of the Virtual User D of structure and the user characteristics of the first child user and the second child user at cut off.Such as, the first child user is 20 years old, and the second child user is 30 years old, and cut off B is being positioned at the centre position of line segment A, and the age of the Virtual User of structure can be 25 years old.Usually, when concrete realization, the general line segment A that selects is less than the line segment of specifying line segment length.What four-headed arrow represented in figure 3 is described line segment A.
Certainly in concrete implementation procedure, also have the user characteristics of a lot of determination Virtual User, build the method for Virtual User, have and realize easy feature.
Embodiment four:
As shown in Figure 1, the present embodiment provides a kind of targeted customer's orientation method, and described method comprises: step S110: based on the user characteristics of seed user, generating virtual user; Here seed user can be the user of advertiser by providing, also can for the user determined according to the history ad data of advertisement; Step S120: using described seed user and described Virtual User jointly as positive sample of users; Like this, the expansion of positive sample of users is just achieved by the generation of Virtual User; Step S130: determine negative sample user; Such as can user beyond Stochastic choice seed user as described negative sample user, certain customers also can be selected from negative sample user as described negative sample user; Step S140: the user characteristics extracting described positive sample of users and described negative sample user respectively; Step S150: based on described orientation parameter, orients the targeted customer meeting predetermined similarity condition with described seed user.
Described step S130 can comprise in the present embodiment: based on the user characteristics of described seed user, calculates the similarity of alternative sample of users and described seed user; Based on described similarity, determine the described alternative sample of users meeting negative sample user condition, as the negative sample user in described sample of users.
In the prior art when localizing objects user, negative sample user is the user of Stochastic choice, this just may cause occurring seed user in negative sample user or having a lot of similarity with seed user, obviously like this targeted customer can be caused to locate coarse phenomenon.In addition, also all users beyond seed user may be considered as negative sample user in the prior art, obviously so positive sample of users can be caused far less than the phenomenon of negative sample user, just there will be the data nonbalance between positive and negative sample of users.In the present embodiment, on the one hand by the generation of Virtual User, carry out the expansion of the number of positive sample of users, also the user characteristics based on seed user is screened negative sample user from alternative user simultaneously, obviously also can reduce the quantity of negative sample user like this.
In the present embodiment, utilize the user characteristics of seed user in the present embodiment, calculate the similarity of alternative sample of users and seed user, here Similarity Measure can adopt various similarity calculating method, these similarity calculating methods see prior art, just can not listed at this one by one in detail.
After calculating similarity, select with the enough little alternative sample of users of seed user similarity as described negative sample user.Such as, selecting similarity to be less than specifies the user of threshold value as described negative sample user.Here appointment threshold value is generally a less value.For another example, described similarity is carried out sequence from small to large, the alternative sample of users of the m position that selected and sorted is forward is as described negative sample user.
Fig. 5 A and Fig. 5 B is the user characteristics utilizing seed user described in the present embodiment, determines the effect schematic diagram of negative sample user.Because seed user is positive sample of users; Obtain negative sample user based on Similarity Measure in the present embodiment, the number of the N number of sample of users obviously obtained based on the user characteristics of seed user is the number M's being greater than seed user.
Negative sample user is determined by the mode of Similarity Measure in the present embodiment, can ensure that the difference between positive sample of users and negative sample user is enough large like this, thus reduce because very similar between positive sample of users and negative sample user, the phenomenons such as the targeted customer's setting accuracy caused is inadequate, and the quantity that targeted customer locates is few.
Embodiment five:
As shown in Figure 1, the present embodiment provides a kind of targeted customer's orientation method, and described method comprises: step S110: based on the user characteristics of seed user, generating virtual user; Here seed user can be the user of advertiser by providing, also can for the user determined according to the history ad data of advertisement; Step S120: using described seed user and described Virtual User jointly as positive sample of users; Like this, the expansion of positive sample of users is just achieved by the generation of Virtual User; Step S130: determine negative sample user; Such as can user beyond Stochastic choice seed user as described negative sample user, certain customers also can be selected from negative sample user as described negative sample user; Step S140: the user characteristics extracting described positive sample of users and described negative sample user respectively; Described step S140 can comprise: utilize the user characteristics of described positive sample of users and described negative sample user to carry out model training, determine the disaggregated model of select target user; Step S150: based on described orientation parameter, orients the targeted customer meeting predetermined similarity condition with described seed user.
In the present embodiment, utilize the model training that the user characteristics of sample of users carries out, the disaggregated model selecting targeted customer and non-targeted user from a large number of users that model training here obtains.The disaggregated model that obtains of model training, can comprise the location algorithm carrying out targeted customer location in the present embodiment, and the information such as each known quantity in location algorithm.
As shown in Figure 6, step S140 described in the present embodiment specifically can comprise:
Step S141: described positive sample of users and negative sample user are divided into training set, checking collection and test set;
Step S142: utilize described training set, adopts different training algorithms to carry out model training;
Step S143: utilize described checking to collect checking to described model training the need of continuation;
Step S144: after the described model training of stopping, utilizing described test set to carry out recruitment evaluation to the alternative model that each described training algorithm obtains;
Step S145: based on described recruitment evaluation, selects a described alternative model as described disaggregated model.
Always sample of users is divided into three set at the present embodiment; These three set are training set, checking collection and test set respectively.Sample of users in step s 141 comprises positive sample of users and negative sample user.These three set are chosen as and include positive sample of users and negative sample user.Described training set, checking in the present embodiment collects different with the user included by any two of test set, to guarantee to obtain optimal disaggregated model.
Utilizing training set to carry out model training to training calculation in each, determining that each training algorithm whether can deconditioning at utilization checking collection.Like this, if utilize 10 training algorithms to carry out model training, trained by training set and pass through checking collection checking after, will the alternative model that 10 can be carried out the disaggregated model of users classification be obtained.To test set be utilized to carry out recruitment evaluation to each alternative model in the present embodiment, select the highest alternative model of degree of accuracy as described disaggregated model, like this, can guarantee accurately to locate described targeted customer.
Such as, utilizing training set to obtaining an alternative model A, utilizing checking to concentrate user characteristics to be input to alternative model A; Alternative model A concentrates the user characteristics of each user to carry out information processing to checking, the result obtained determines that this user is positive sample of users or negative sample user, again the data of result and sample of users are compared, obviously just can determine the correct probability of result.If the correct probability of the result of alternative model A exceedes appointment probability, then think that the training of alternative model A can stopped.
The user characteristics of sample of users in test set will be input in alternative model A in step S144, reach the correct probability of alternative model A.If now, also have alternative model B and alternative model C, can by comparing the correct probability that this three alternative model obtain in step S134 in step S135, select correct probability the highest as final disaggregated model.
Thus model training described in the present embodiment not only can determine the known quantity of carrying out targeted customer location, also position suitable location algorithm by determining, can the accurate location of realize target user.
Such as, even the equipment that various study neural networks etc. can be utilized in the present embodiment to have learning ability adopts LR-FTRL algorithm and the algorithm scheduling algorithm based on artificial regular Rule-Based, described LR-FTRL algorithm specifically can be utilized to select and to get LR model continuously; Utilize the algorithm of described Rule-Based can obtain RuleSet model.
LR in described LR-FTRL algorithm is the abbreviation of LogisticRegressionlogistic, and corresponding Chinese regression model, is a kind of linear regression analysis model of broad sense, is usually used in data mining, disease automatic diagnosis, the fields such as economic projection.FTRL in described LR-FTRL algorithm is the abbreviation of FollowtheRegularizedLeader.LR-FTRL a kind of on-line optimization regression model algorithm at last, supports the learning rate of each characteristic dimension that each dimension is considered in a word, has good convergence and the fast feature of training speed.
There is provided a kind of targeted customer's orientation method below in conjunction with the present embodiment, as shown in Figure 7, based on the training schematic flow sheet that method described in the present embodiment provides, comprise four parts:
Part I:
Carry out labeled data structure.Here labeled data builds to comprise determines positive sample of users and negative sample user.
Seed user is utilized to carry out all more than the seed user positive sample of users of over-sampling process.Here card sample of users can comprise the Virtual User that the user characteristics based on seed user is determined, also can comprise the seed user etc. that the expansion based on the seed user of the similar advertisement of the advertisement corresponding with seed user obtains.
Carry out lack sampling to alternative user and carry out user's selection, obtain negative sample user, the quantity of negative sample user is here fewer than the quantity of user selected in Fig. 7.
Part II:
Obtain the user characteristics of sample of users.Safeguard there is the various feature of the user of user in a database, the user's portrait such as formed based on age of user, sex, hobby, moving position and education background etc., the social networks of acquisition user, the advertisement behavior carried out for advertisement, the App behavior that utilizes App to carry out; Carry out feature extraction, carry out feature normalization and sliding-model control obtains user characteristics.Here normalized utilizes the unified of the feature name characterizing same information to wait operation.
Part III:
Disaggregated model is determined.Here the determination of disaggregated model comprises the data obtaining positive sample of users and negative sample user, the user characteristics being extracted sample of users; Sample of users is divided in order to training set, checking collection and test set three set, utilize different training algorithms to carry out training and obtain training pattern, utilize checking collection to determine when the training of deconditioning model, utilize test set to train the training pattern obtained to select one as final disaggregated model from multiple algorithms of different.
Part IV:
Online user is directed.Here namely online user's orientation comprises the location of targeted customer.First, extract the user characteristics of whole user, utilize disaggregated model to carry out users classification and estimate, the parameter of the measurement user obtained and the similarity of seed user.Then select similarity the user of top N as targeted customer, remaining like this user is non-targeted user, just completes users classification.The most backward targeted customer throws in corresponding advertisement, thus promotes advertisement delivery effect, promotes the conversion ratio of advertisement.
Embodiment six:
As shown in Figure 8, the present embodiment provides a kind of targeted customer's orienting device, and described device comprises:
Generation unit 110, for the user characteristics based on seed user, generating virtual user;
First determining unit 120, for using described seed user and described Virtual User jointly as positive sample of users;
Second determining unit 130, for determining negative sample user;
Training unit 140, for extracting the user characteristics of described positive sample of users and described negative sample user respectively;
Positioning unit 150, for based on described orientation parameter, orients the targeted customer meeting predetermined similarity condition with described seed user.
Targeted customer's orienting device described in the present embodiment can be and is applied to the various device with the electronic equipment of information processing, utilizes the device being applied to the server of social platform or the processor of advertisement launching platform.
The concrete structure of described generation unit 110, first determining unit 120, training unit 140 and positioning unit 150 may correspond in processor or treatment circuit.Described processor can comprise central processing unit, microprocessor, digital signal processor or programmable array etc.Described treatment circuit can comprise special IC etc.
Described generation unit 110, first determining unit 120, training unit 140 and positioning unit 150 are integrated corresponds to same processor or treatment circuit, also can correspond respectively to different processors or treatment circuit.When corresponding to same processor at least two of generation unit 110, first determining unit 120, training unit 140 and positioning unit 150, this processor can adopt the mode of time division multiplex or concurrent thread to complete the function of described generation unit 110, first determining unit 120, training unit 140 and positioning unit 150 respectively.Such as, described processor or treatment circuit, by performing predetermined instruction, realize the function of above-mentioned generation unit 110, first determining unit 120, training unit 140 and positioning unit 150.
As shown in Figure 9, the present embodiment also provides another kind of described targeted customer's orienting device, and described device comprises processor 220, storage medium 240, at least one external communication interface 210 and display screen 250; Described processor 220, storage medium 304 and external communication interface 301 are all connected by bus 303.Described storage medium 240 stores computer executable instructions; Described processor 220 performs the described computer executable instructions stored in described storage medium 304, at least can realize: based on the user characteristics of seed user, generating virtual user; Using described seed user and described Virtual User jointly as positive sample of users; Determine negative sample user; Extract the user characteristics of described positive sample of users and described negative sample user respectively; User characteristics based on described positive sample of users and described negative sample user is trained, and determines orientation parameter; Based on described orientation parameter, orient the targeted customer meeting predetermined similarity condition with described seed user.
In a word, device described in the present embodiment, can based on the user characteristics of child user, generating virtual user, again using seed user and Virtual User jointly as positive sample of users, which adds the number of positive sample of users, positive number of samples and negative sample number can be reduced, the orientation parameter that the training that imbalance causes obtains is strong not to the distinguishing ability of targeted customer, causes targeted customer to locate coarse feature.
Embodiment seven:
As shown in Figure 8, the present embodiment provides a kind of targeted customer's orienting device, and described device comprises: generation unit 110, for the user characteristics based on seed user, and generating virtual user; Here described generation unit 110, can specifically for determining the primary importance of described seed user at feature space; Based at least two described primary importance determination second places; The user characteristics of described Virtual User is determined and generating virtual user based on the described second place; First determining unit 120, for using described seed user and described Virtual User jointly as positive sample of users; Second determining unit 130, for determining negative sample user; Training unit 140, for extracting the user characteristics of described positive sample of users and described negative sample user respectively; Positioning unit 150, for based on described orientation parameter, orients the targeted customer meeting predetermined similarity condition with described seed user.
The hardware configuration of generation unit 110 correspondence described in the present embodiment can see in previous embodiment, and described generation unit 110 is by morphogenesis characters space in the present embodiment, and determines the primary importance of seed user at feature space; Determine the second place according to the primary importance of at least two seed user again, determine the user characteristics of described Virtual User based on the second place in the value of user characteristics corresponding to feature space, thus complete the generation of Virtual User; Have structure simple, realize the large feature of the similarity of Virtual User that is simple and that build and seed user.
In a word, present embodiments provide a kind of device, can according to the user characteristics of seed user, form Virtual User, and using Virtual User and seed user jointly as positive sample of users.Described like this training unit 140 just can have more sample data, thus can the orientation parameter of more accurate localizing objects user by obtaining.
Embodiment eight:
As shown in Figure 8, the present embodiment provides a kind of targeted customer's orienting device, and described device comprises: generation unit 110, for the user characteristics based on seed user, and generating virtual user; Here described generation unit, specifically for extracting the numerical value user characteristics in the user characteristics of described seed user; Based on described numerical value user characteristics, determine the range of choices of numerical value user characteristics; The numerical value user characteristics of a numerical value as described Virtual User is selected from described range of choices; Extract the nonumeric user characteristics in the user characteristics of described seed user, and give probable value to described nonumeric user characteristics; The nonumeric user characteristics of described Virtual User is determined based on described probable value; First determining unit 120, for using described seed user and described Virtual User jointly as positive sample of users; Second determining unit 130, for determining negative sample user; Training unit 140, for extracting the user characteristics of described positive sample of users and described negative sample user respectively; Positioning unit 150, for based on described orientation parameter, orients the targeted customer meeting predetermined similarity condition with described seed user.
The basis of the targeted customer's orienting device described in the present embodiment is improved further, the hardware configuration of described generation unit 110 correspondence also can see in previous embodiment in the present embodiment, just user characteristics will be divided into numerical value user characteristics and nonumeric user characteristics by described generation unit 110 in the present embodiment, the user characteristics of Virtual User is built again respectively according to the feature of the nonumeric user characteristics of numerical value user characteristics, thus complete the generation of Virtual User, there is structure equally simple and realize easy feature.
In a word, present embodiments provide a kind of device, can according to the user characteristics of seed user, form Virtual User, and using Virtual User and seed user jointly as positive sample of users.Described like this training unit 140 just can have more sample data, thus can the orientation parameter of more accurate localizing objects user by obtaining.
Embodiment nine:
As shown in Figure 8, the present embodiment provides a kind of targeted customer's orienting device, and described device comprises: generation unit 110, for the user characteristics based on seed user, and generating virtual user; First determining unit 120, for using described seed user and described Virtual User jointly as positive sample of users; Second determining unit 130, for determining negative sample user; Here described second determining unit 130, specifically for the user characteristics based on described seed user, calculates the similarity of alternative sample of users and described seed user; Based on described similarity, determine the described alternative sample of users meeting negative sample user condition, as the negative sample user in described sample of users; Training unit 140, for extracting the user characteristics of described positive sample of users and described negative sample user respectively; Positioning unit 150, for based on described orientation parameter, orients the targeted customer meeting predetermined similarity condition with described seed user.
Negative sample user is selected in the present embodiment by according to the user characteristics of seed user.Here negative sample user is the alternative user less with the similarity of seed user, such as similarity is less than the user of predetermined threshold value, or the alternative user to rank behind with the similarity of seed user, the similarity of these users and seed user is little, obviously can watch same advertisement or perform the probability of same operation low, therefore the negative sample user of orientation parameter can be obtained as training.The concrete structure of generation unit 110 described in the present embodiment can see previous embodiment, but generation unit 110 has expanded out negative sample user based on the user characteristics of seed user in the present embodiment, relative to random selecting user as negative sample user, the phenomenon that the degree of accuracy that the user very similar to seed user caused as negative sample user is inadequate obviously can be reduced.
Embodiment ten:
As shown in Figure 8, the present embodiment provides a kind of targeted customer's orienting device, and described device comprises: generation unit 110, for the user characteristics based on seed user, and generating virtual user; First determining unit 120, for using described seed user and described Virtual User jointly as positive sample of users; Second determining unit 130, for determining negative sample user; Training unit 140, for extracting the user characteristics of described positive sample of users and described negative sample user respectively; Here described training unit 140, specifically for utilizing the user characteristics of described positive sample of users and described negative sample user to carry out model training, determines the disaggregated model of select target user; Positioning unit 150, for based on described orientation parameter, orients the targeted customer meeting predetermined similarity condition with described seed user.
In a word, the orientation parameter that described in the present embodiment, training unit 140 finally obtains is a systematic disaggregated model, follow-up from a large amount of users during select target user, by running this disaggregated model, user can be divided into targeted customer and non-targeted user, having can the feature of accurate localizing objects user.
As further improvement of this embodiment, described training unit 140, specifically for being divided into training set, checking collection and test set by described positive sample of users and negative sample user; Utilize described training set, adopt different training algorithms to carry out model training; Described checking is utilized to collect checking to described model training the need of continuation; After the described model training of stopping, described test set is utilized to carry out recruitment evaluation to the alternative model that each described training algorithm obtains; Based on described recruitment evaluation, select a described alternative model as described disaggregated model.
In the present embodiment sample of users is divided in order to three set, training set, checking collection and test set respectively, positive sample of users and negative sample user in each set, the training of the alternative model of disaggregated model is carried out by training set, determine whether to need to continue training by checking collection, by test set from the optimum training pattern of selection one as final disaggregated model, the training pattern obtained so further can promote the degree of accuracy of targeted customer's orientation.
The concrete structure of described training unit 140 to corresponding to the structure such as various learning machine or neural network, by the input of the data of sample of users, can be trained and obtain described disaggregated model in the present embodiment.At this disaggregated model of user, other users are classified, just can be easy and accurately filter out described targeted customer.
In several embodiments that the application provides, should be understood that disclosed equipment and method can realize by another way.Apparatus embodiments described above is only schematic, such as, the division of described unit, be only a kind of logic function to divide, actual can have other dividing mode when realizing, and as: multiple unit or assembly can be in conjunction with, maybe can be integrated into another system, or some features can be ignored, or do not perform.In addition, the coupling each other of shown or discussed each ingredient or direct-coupling or communication connection can be by some interfaces, and the indirect coupling of equipment or unit or communication connection can be electrical, machinery or other form.
The above-mentioned unit illustrated as separating component or can may not be and physically separates, and the parts as unit display can be or may not be physical location, namely can be positioned at a place, also can be distributed in multiple network element; Part or all of unit wherein can be selected according to the actual needs to realize the object of the present embodiment scheme.
In addition, each functional unit in various embodiments of the present invention can all be integrated in a processing module, also can be each unit individually as a unit, also can two or more unit in a unit integrated; Above-mentioned integrated unit both can adopt the form of hardware to realize, and the form that hardware also can be adopted to add SFU software functional unit realizes.
One of ordinary skill in the art will appreciate that: all or part of step realizing said method embodiment can have been come by the hardware that programmed instruction is relevant, aforesaid program can be stored in a computer read/write memory medium, this program, when performing, performs the step comprising said method embodiment; And aforesaid storage medium comprises: movable storage device, ROM (read-only memory) (ROM, Read-OnlyMemory), random access memory (RAM, RandomAccessMemory), magnetic disc or CD etc. various can be program code stored medium.
The above; be only the specific embodiment of the present invention, but protection scope of the present invention is not limited thereto, is anyly familiar with those skilled in the art in the technical scope that the present invention discloses; change can be expected easily or replace, all should be encompassed within protection scope of the present invention.Therefore, protection scope of the present invention should be as the criterion with the protection domain of described claim.
Claims (12)
1. targeted customer's orientation method, is characterized in that, described method comprises:
Based on the user characteristics of seed user, generating virtual user;
Using described seed user and described Virtual User jointly as positive sample of users;
Determine negative sample user;
Extract the user characteristics of described positive sample of users and described negative sample user respectively;
User characteristics based on described positive sample of users and described negative sample user is trained, and determines orientation parameter;
Based on described orientation parameter, orient the targeted customer meeting predetermined similarity condition with described seed user.
2. method according to claim 1, is characterized in that,
The described user characteristics based on seed user, generating virtual user, comprising:
Determine the primary importance of described seed user at feature space;
Based at least two described primary importance determination second places;
The user characteristics of described Virtual User is determined and generating virtual user based on the described second place.
3. method according to claim 1, is characterized in that,
The described user characteristics based on seed user, generating virtual user, comprising:
Extract the numerical value user characteristics in the user characteristics of described seed user;
Based on described numerical value user characteristics, determine the range of choices of numerical value user characteristics;
The numerical value user characteristics of a numerical value as described Virtual User is selected from described range of choices;
Extract the nonumeric user characteristics in the user characteristics of described seed user, and give probable value to described nonumeric user characteristics;
The nonumeric user characteristics of described Virtual User is determined based on described probable value.
4. method according to claim 1, is characterized in that,
Describedly determine negative sample user, comprising:
Based on the user characteristics of described seed user, calculate the similarity of alternative sample of users and described seed user;
Based on described similarity, determine the described alternative sample of users meeting negative sample user condition, as the negative sample user in described sample of users.
5. the method according to any one of Claims 1-4, is characterized in that,
The described user characteristics based on described positive sample of users and described negative sample user is trained, and determines orientation parameter, comprising:
Utilize the user characteristics of described positive sample of users and described negative sample user to carry out model training, determine the disaggregated model of select target user.
6. method according to claim 5, is characterized in that,
The described user characteristics of described sample of users that utilizes carries out model training, determines the training pattern of select target user, comprising:
Described positive sample of users and negative sample user are divided into training set, checking collection and test set;
Utilize described training set, adopt different training algorithms to carry out model training;
Described checking is utilized to collect checking to described model training the need of continuation;
After the described model training of stopping, described test set is utilized to carry out recruitment evaluation to the alternative model that each described training algorithm obtains;
Based on described recruitment evaluation, select a described alternative model as described disaggregated model.
7. targeted customer's orienting device, is characterized in that, described device comprises:
Generation unit, for the user characteristics based on seed user, generating virtual user;
First determining unit, for using described seed user and described Virtual User jointly as positive sample of users;
Second determining unit, for determining negative sample user;
Training unit, for extracting the user characteristics of described positive sample of users and described negative sample user respectively;
Positioning unit, for based on described orientation parameter, orients the targeted customer meeting predetermined similarity condition with described seed user.
8. device according to claim 7, is characterized in that,
Described generation unit, for determining the primary importance of described seed user at feature space; Based at least two described primary importance determination second places; The user characteristics of described Virtual User is determined and generating virtual user based on the described second place.
9. device according to claim 7, is characterized in that,
Described generation unit, specifically for extracting the numerical value user characteristics in the user characteristics of described seed user; Based on described numerical value user characteristics, determine the range of choices of numerical value user characteristics; The numerical value user characteristics of a numerical value as described Virtual User is selected from described range of choices; Extract the nonumeric user characteristics in the user characteristics of described seed user, and give probable value to described nonumeric user characteristics; The nonumeric user characteristics of described Virtual User is determined based on described probable value.
10. device according to claim 7, is characterized in that,
Described second determining unit, specifically for the user characteristics based on described seed user, calculates the similarity of alternative sample of users and described seed user; Based on described similarity, determine the described alternative sample of users meeting negative sample user condition, as the negative sample user in described sample of users.
11. devices according to any one of claim 7 to 10, is characterized in that,
Described training unit, specifically for utilizing the user characteristics of described positive sample of users and described negative sample user to carry out model training, determines the disaggregated model of select target user.
12. devices according to claim 11, is characterized in that,
Described training unit, specifically for being divided into training set, checking collection and test set by described positive sample of users and negative sample user; Utilize described training set, adopt different training algorithms to carry out model training; Described checking is utilized to collect checking to described model training the need of continuation; After the described model training of stopping, described test set is utilized to carry out recruitment evaluation to the alternative model that each described training algorithm obtains; Based on described recruitment evaluation, select a described alternative model as described disaggregated model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510996108.5A CN105447730B (en) | 2015-12-25 | 2015-12-25 | Target user orientation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510996108.5A CN105447730B (en) | 2015-12-25 | 2015-12-25 | Target user orientation method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105447730A true CN105447730A (en) | 2016-03-30 |
| CN105447730B CN105447730B (en) | 2020-11-06 |
Family
ID=55557866
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510996108.5A Active CN105447730B (en) | 2015-12-25 | 2015-12-25 | Target user orientation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105447730B (en) |
Cited By (44)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106204063A (en) * | 2016-06-30 | 2016-12-07 | 北京奇艺世纪科技有限公司 | A kind of paying customer's method for digging and device |
| CN106357517A (en) * | 2016-09-27 | 2017-01-25 | 腾讯科技（北京）有限公司 | Directional label generation method and device |
| CN106355449A (en) * | 2016-08-31 | 2017-01-25 | 腾讯科技（深圳）有限公司 | User selecting method and device |
| CN106408411A (en) * | 2016-08-31 | 2017-02-15 | 北京城市网邻信息技术有限公司 | Credit assessment method and device |
| CN107292342A (en) * | 2017-06-21 | 2017-10-24 | 广东欧珀移动通信有限公司 | Data processing method and related product |
| WO2017190610A1 (en) * | 2016-05-05 | 2017-11-09 | 腾讯科技（深圳）有限公司 | Target user orientation method and device, and computer storage medium |
| CN107346448A (en) * | 2016-05-06 | 2017-11-14 | 富士通株式会社 | Identification device, trainer and method based on deep neural network |
| CN107506402A (en) * | 2017-08-03 | 2017-12-22 | 北京百度网讯科技有限公司 | Sort method, device, equipment and the computer-readable recording medium of search result |
| 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 |
| CN107590691A (en) * | 2017-09-06 | 2018-01-16 | 晶赞广告（上海）有限公司 | A kind of information issuing method and device, storage medium, terminal |
| CN107679920A (en) * | 2017-10-20 | 2018-02-09 | 北京奇艺世纪科技有限公司 | The put-on method and device of a kind of advertisement |
| CN108038739A (en) * | 2017-12-27 | 2018-05-15 | 北京奇虎科技有限公司 | A kind of method and system that extending user is determined according to the statistics degree of association |
| CN108053260A (en) * | 2017-12-27 | 2018-05-18 | 北京奇虎科技有限公司 | A kind of method and system that extending user is determined according to statistics interest-degree |
| CN108122123A (en) * | 2016-11-29 | 2018-06-05 | 华为技术有限公司 | A kind of method and device for extending potential user |
| CN108182600A (en) * | 2017-12-27 | 2018-06-19 | 北京奇虎科技有限公司 | A kind of method and system that extending user is determined according to weighted calculation |
| CN108415913A (en) * | 2017-02-09 | 2018-08-17 | 周孟 | Crowd's orientation method based on uncertain neighbours |
| CN108427690A (en) * | 2017-02-15 | 2018-08-21 | 腾讯科技（深圳）有限公司 | Information distribution method and device |
| CN108460396A (en) * | 2017-09-20 | 2018-08-28 | 腾讯科技（深圳）有限公司 | The negative method of sampling and device |
| CN108647983A (en) * | 2018-03-16 | 2018-10-12 | 北京奇艺世纪科技有限公司 | Seed user determines method, apparatus and advertisement placement method, device |
| CN108647986A (en) * | 2018-03-28 | 2018-10-12 | 北京奇艺世纪科技有限公司 | A kind of target user determines method, apparatus and electronic equipment |
| CN108648093A (en) * | 2018-04-23 | 2018-10-12 | 腾讯科技（深圳）有限公司 | Data processing method, device and equipment |
| CN108647990A (en) * | 2018-04-04 | 2018-10-12 | 北京奇艺世纪科技有限公司 | A kind of method, apparatus and electronic equipment of determining target user |
| WO2018192348A1 (en) * | 2017-04-20 | 2018-10-25 | 腾讯科技（深圳）有限公司 | Data processing method and device, and server |
| CN108734304A (en) * | 2018-05-31 | 2018-11-02 | 阿里巴巴集团控股有限公司 | A kind of training method of data model, device and computer equipment |
| CN109087162A (en) * | 2018-07-05 | 2018-12-25 | 杭州朗和科技有限公司 | Data processing method, system, medium and calculating equipment |
| CN109242537A (en) * | 2018-08-14 | 2019-01-18 | 平安普惠企业管理有限公司 | Advertisement placement method, device, computer equipment and storage medium |
| CN109255656A (en) * | 2018-08-31 | 2019-01-22 | 有米科技股份有限公司 | A kind of user's extended method, apparatus and system based on composite model |
| CN109446432A (en) * | 2018-12-17 | 2019-03-08 | 微梦创科网络科技（中国）有限公司 | Method and device for recommending information |
| CN110008973A (en) * | 2018-11-23 | 2019-07-12 | 阿里巴巴集团控股有限公司 | A kind of model training method, the method and device that target user is determined based on model |
| CN110097395A (en) * | 2019-03-27 | 2019-08-06 | 平安科技（深圳）有限公司 | Directional advertisement release method, device and computer readable storage medium |
| CN110147882A (en) * | 2018-09-03 | 2019-08-20 | 腾讯科技（深圳）有限公司 | Training method, crowd's method of diffusion, device and the equipment of neural network model |
| CN110163747A (en) * | 2019-05-24 | 2019-08-23 | 同盾控股有限公司 | Target recommended method for digging, device, medium and electronic equipment |
| WO2019169961A1 (en) * | 2018-03-06 | 2019-09-12 | 阿里巴巴集团控股有限公司 | Method and device for determining group of target users |
| CN110245687A (en) * | 2019-05-17 | 2019-09-17 | 腾讯科技（上海）有限公司 | User classification method and device |
| CN111126614A (en) * | 2018-11-01 | 2020-05-08 | 百度在线网络技术（北京）有限公司 | Attribution method, attribution device and storage medium |
| CN111131356A (en) * | 2018-10-31 | 2020-05-08 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating information |
| CN111275205A (en) * | 2020-01-13 | 2020-06-12 | 优地网络有限公司 | Virtual sample generation method, terminal device and storage medium |
| CN111444930A (en) * | 2019-01-17 | 2020-07-24 | 上海游昆信息技术有限公司 | Method and device for determining prediction effect of two-classification model |
| WO2020228514A1 (en) * | 2019-05-13 | 2020-11-19 | 腾讯科技（深圳）有限公司 | Content recommendation method and apparatus, and device and storage medium |
| CN112925973A (en) * | 2019-12-06 | 2021-06-08 | 北京沃东天骏信息技术有限公司 | Data processing method and device |
| CN112989213A (en) * | 2021-05-19 | 2021-06-18 | 腾讯科技（深圳）有限公司 | Content recommendation method, device and system, electronic equipment and storage medium |
| CN112989179A (en) * | 2019-12-13 | 2021-06-18 | 北京达佳互联信息技术有限公司 | Model training and multimedia content recommendation method and device |
| CN113536848A (en) * | 2020-04-17 | 2021-10-22 | 中国移动通信集团广东有限公司 | Data processing method and device and electronic equipment |
| CN115545779A (en) * | 2022-10-11 | 2022-12-30 | 西窗科技(苏州)有限公司 | Big data-based advertisement delivery early warning management method and system |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103678329B (en) * | 2012-09-04 | 2018-05-04 | 中兴通讯股份有限公司 | Recommend method and device |
| CN111325416A (en) * | 2014-12-09 | 2020-06-23 | 北京嘀嘀无限科技发展有限公司 | Method and device for predicting user loss of taxi calling platform |
| CN105046320A (en) * | 2015-08-13 | 2015-11-11 | 中国人民解放军61599部队计算所 | Virtual sample generation method |
-
2015
- 2015-12-25 CN CN201510996108.5A patent/CN105447730B/en active Active
Cited By (73)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017190610A1 (en) * | 2016-05-05 | 2017-11-09 | 腾讯科技（深圳）有限公司 | Target user orientation method and device, and computer storage medium |
| US11004107B2 (en) | 2016-05-05 | 2021-05-11 | Tencent Technology (Shenzhen) Company Limited | Target user directing method and apparatus and computer storage medium |
| CN107346496B (en) * | 2016-05-05 | 2021-12-10 | 腾讯科技（北京）有限公司 | Target user orientation method and device |
| EP3407285B1 (en) * | 2016-05-05 | 2023-06-28 | Tencent Technology (Shenzhen) Company Limited | Target user orientation method and device, and computer storage medium |
| CN107346496A (en) * | 2016-05-05 | 2017-11-14 | 腾讯科技（北京）有限公司 | Targeted customer's orientation method and device |
| CN107346448A (en) * | 2016-05-06 | 2017-11-14 | 富士通株式会社 | Identification device, trainer and method based on deep neural network |
| CN107346448B (en) * | 2016-05-06 | 2021-12-21 | 富士通株式会社 | Deep neural network-based recognition device, training device and method |
| US11049007B2 (en) | 2016-05-06 | 2021-06-29 | Fujitsu Limited | Recognition apparatus based on deep neural network, training apparatus and methods thereof |
| CN106204063A (en) * | 2016-06-30 | 2016-12-07 | 北京奇艺世纪科技有限公司 | A kind of paying customer's method for digging and device |
| CN106355449B (en) * | 2016-08-31 | 2021-09-07 | 腾讯科技（深圳）有限公司 | User selection method and device |
| CN106408411A (en) * | 2016-08-31 | 2017-02-15 | 北京城市网邻信息技术有限公司 | Credit assessment method and device |
| US11574139B2 (en) | 2016-08-31 | 2023-02-07 | Tencent Technology (Shenzhen) Company Limited | Information pushing method, storage medium and server |
| CN106355449A (en) * | 2016-08-31 | 2017-01-25 | 腾讯科技（深圳）有限公司 | User selecting method and device |
| WO2018059176A1 (en) * | 2016-09-27 | 2018-04-05 | 腾讯科技（深圳）有限公司 | Method and apparatus for generating targeted label and storage medium |
| US11270343B2 (en) | 2016-09-27 | 2022-03-08 | Tencent Technology (Shenzhen) Company Limited | Method and apparatus for generating targeted label, and storage medium |
| CN106357517B (en) * | 2016-09-27 | 2020-09-11 | 腾讯科技（北京）有限公司 | Directional label generation method and device |
| CN106357517A (en) * | 2016-09-27 | 2017-01-25 | 腾讯科技（北京）有限公司 | Directional label generation method and device |
| CN108122123A (en) * | 2016-11-29 | 2018-06-05 | 华为技术有限公司 | A kind of method and device for extending potential user |
| WO2018099177A1 (en) * | 2016-11-29 | 2018-06-07 | 华为技术有限公司 | Potential user expansion method and device |
| CN108122123B (en) * | 2016-11-29 | 2021-08-20 | 华为技术有限公司 | A method and apparatus for expanding potential users |
| CN108415913A (en) * | 2017-02-09 | 2018-08-17 | 周孟 | Crowd's orientation method based on uncertain neighbours |
| CN108427690A (en) * | 2017-02-15 | 2018-08-21 | 腾讯科技（深圳）有限公司 | Information distribution method and device |
| WO2018192348A1 (en) * | 2017-04-20 | 2018-10-25 | 腾讯科技（深圳）有限公司 | Data processing method and device, and server |
| CN107292342A (en) * | 2017-06-21 | 2017-10-24 | 广东欧珀移动通信有限公司 | Data processing method and related product |
| CN107506402A (en) * | 2017-08-03 | 2017-12-22 | 北京百度网讯科技有限公司 | Sort method, device, equipment and the computer-readable recording medium of search result |
| CN107590691A (en) * | 2017-09-06 | 2018-01-16 | 晶赞广告（上海）有限公司 | A kind of information issuing method and device, storage medium, terminal |
| CN107590691B (en) * | 2017-09-06 | 2021-01-15 | 晶赞广告（上海）有限公司 | Information publishing method and device, storage medium and terminal |
| CN108460396B (en) * | 2017-09-20 | 2021-10-15 | 腾讯科技（深圳）有限公司 | Negative sampling method and device |
| CN108460396A (en) * | 2017-09-20 | 2018-08-28 | 腾讯科技（深圳）有限公司 | The negative method of sampling and device |
| WO2019061905A1 (en) * | 2017-09-28 | 2019-04-04 | 北京小度信息科技有限公司 | User behavior prediction method and apparatus, and electronic device |
| 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 |
| CN107679920A (en) * | 2017-10-20 | 2018-02-09 | 北京奇艺世纪科技有限公司 | The put-on method and device of a kind of advertisement |
| CN108038739A (en) * | 2017-12-27 | 2018-05-15 | 北京奇虎科技有限公司 | A kind of method and system that extending user is determined according to the statistics degree of association |
| CN108053260A (en) * | 2017-12-27 | 2018-05-18 | 北京奇虎科技有限公司 | A kind of method and system that extending user is determined according to statistics interest-degree |
| CN108182600A (en) * | 2017-12-27 | 2018-06-19 | 北京奇虎科技有限公司 | A kind of method and system that extending user is determined according to weighted calculation |
| WO2019169961A1 (en) * | 2018-03-06 | 2019-09-12 | 阿里巴巴集团控股有限公司 | Method and device for determining group of target users |
| CN108647983A (en) * | 2018-03-16 | 2018-10-12 | 北京奇艺世纪科技有限公司 | Seed user determines method, apparatus and advertisement placement method, device |
| CN108647986A (en) * | 2018-03-28 | 2018-10-12 | 北京奇艺世纪科技有限公司 | A kind of target user determines method, apparatus and electronic equipment |
| CN108647990A (en) * | 2018-04-04 | 2018-10-12 | 北京奇艺世纪科技有限公司 | A kind of method, apparatus and electronic equipment of determining target user |
| CN108648093A (en) * | 2018-04-23 | 2018-10-12 | 腾讯科技（深圳）有限公司 | Data processing method, device and equipment |
| CN108648093B (en) * | 2018-04-23 | 2021-11-09 | 腾讯科技（深圳）有限公司 | Data processing method, device and equipment |
| CN108734304A (en) * | 2018-05-31 | 2018-11-02 | 阿里巴巴集团控股有限公司 | A kind of training method of data model, device and computer equipment |
| CN108734304B (en) * | 2018-05-31 | 2022-04-19 | 创新先进技术有限公司 | Training method and device of data model and computer equipment |
| CN109087162A (en) * | 2018-07-05 | 2018-12-25 | 杭州朗和科技有限公司 | Data processing method, system, medium and calculating equipment |
| CN109242537A (en) * | 2018-08-14 | 2019-01-18 | 平安普惠企业管理有限公司 | Advertisement placement method, device, computer equipment and storage medium |
| CN109255656A (en) * | 2018-08-31 | 2019-01-22 | 有米科技股份有限公司 | A kind of user's extended method, apparatus and system based on composite model |
| CN109255656B (en) * | 2018-08-31 | 2020-09-18 | 有米科技股份有限公司 | User extension method, device and system based on composite model |
| CN110147882B (en) * | 2018-09-03 | 2023-02-10 | 腾讯科技（深圳）有限公司 | Neural network model training method, crowd diffusion method, device and equipment |
| CN110147882A (en) * | 2018-09-03 | 2019-08-20 | 腾讯科技（深圳）有限公司 | Training method, crowd's method of diffusion, device and the equipment of neural network model |
| CN111131356A (en) * | 2018-10-31 | 2020-05-08 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating information |
| CN111126614A (en) * | 2018-11-01 | 2020-05-08 | 百度在线网络技术（北京）有限公司 | Attribution method, attribution device and storage medium |
| CN111126614B (en) * | 2018-11-01 | 2024-01-16 | 百度在线网络技术（北京）有限公司 | Attribution method, attribution device and storage medium |
| CN110008973B (en) * | 2018-11-23 | 2023-05-02 | 创新先进技术有限公司 | Model training method, method and device for determining target user based on model |
| CN110008973A (en) * | 2018-11-23 | 2019-07-12 | 阿里巴巴集团控股有限公司 | A kind of model training method, the method and device that target user is determined based on model |
| CN109446432A (en) * | 2018-12-17 | 2019-03-08 | 微梦创科网络科技（中国）有限公司 | Method and device for recommending information |
| CN111444930A (en) * | 2019-01-17 | 2020-07-24 | 上海游昆信息技术有限公司 | Method and device for determining prediction effect of two-classification model |
| CN110097395A (en) * | 2019-03-27 | 2019-08-06 | 平安科技（深圳）有限公司 | Directional advertisement release method, device and computer readable storage medium |
| CN110097395B (en) * | 2019-03-27 | 2023-05-26 | 平安科技（深圳）有限公司 | Directional advertisement putting method and device and computer readable storage medium |
| US12314832B2 (en) | 2019-05-13 | 2025-05-27 | Tencent Technology (Shenzhen) Company Limited | Content recommendation method and apparatus, device, and storage medium |
| WO2020228514A1 (en) * | 2019-05-13 | 2020-11-19 | 腾讯科技（深圳）有限公司 | Content recommendation method and apparatus, and device and storage medium |
| CN110245687A (en) * | 2019-05-17 | 2019-09-17 | 腾讯科技（上海）有限公司 | User classification method and device |
| CN110245687B (en) * | 2019-05-17 | 2021-06-04 | 腾讯科技（上海）有限公司 | User classification method and device |
| CN110163747A (en) * | 2019-05-24 | 2019-08-23 | 同盾控股有限公司 | Target recommended method for digging, device, medium and electronic equipment |
| CN112925973A (en) * | 2019-12-06 | 2021-06-08 | 北京沃东天骏信息技术有限公司 | Data processing method and device |
| CN112989179A (en) * | 2019-12-13 | 2021-06-18 | 北京达佳互联信息技术有限公司 | Model training and multimedia content recommendation method and device |
| CN112989179B (en) * | 2019-12-13 | 2023-07-28 | 北京达佳互联信息技术有限公司 | Model training and multimedia content recommendation method and device |
| CN111275205B (en) * | 2020-01-13 | 2024-03-22 | 优地网络有限公司 | Virtual sample generation method, terminal equipment and storage medium |
| CN111275205A (en) * | 2020-01-13 | 2020-06-12 | 优地网络有限公司 | Virtual sample generation method, terminal device and storage medium |
| CN113536848A (en) * | 2020-04-17 | 2021-10-22 | 中国移动通信集团广东有限公司 | Data processing method and device and electronic equipment |
| CN113536848B (en) * | 2020-04-17 | 2024-03-19 | 中国移动通信集团广东有限公司 | Data processing method and device and electronic equipment |
| CN112989213A (en) * | 2021-05-19 | 2021-06-18 | 腾讯科技（深圳）有限公司 | Content recommendation method, device and system, electronic equipment and storage medium |
| CN115545779A (en) * | 2022-10-11 | 2022-12-30 | 西窗科技(苏州)有限公司 | Big data-based advertisement delivery early warning management method and system |
| CN115545779B (en) * | 2022-10-11 | 2023-09-01 | 西窗科技(苏州)有限公司 | Early warning management method and system for advertisement delivery based on big data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105447730B (en) | 2020-11-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105447730A (en) | Target user orientation method and device | |
| US11574139B2 (en) | Information pushing method, storage medium and server | |
| US11206311B2 (en) | Method and system for measuring user engagement using click/skip in content stream | |
| EP3779841B1 (en) | Method, apparatus and system for sending information, and computer-readable storage medium | |
| US9706008B2 (en) | Method and system for efficient matching of user profiles with audience segments | |
| CN107346496B (en) | Target user orientation method and device | |
| US10198635B2 (en) | Systems and methods for associating an image with a business venue by using visually-relevant and business-aware semantics | |
| KR101419504B1 (en) | System and method providing a suited shopping information by analyzing the propensity of an user | |
| CN110428298A (en) | A kind of shop recommended method, device and equipment | |
| CN110400169A (en) | A kind of information-pushing method, device and equipment | |
| CN107424043A (en) | A kind of Products Show method and device, electronic equipment | |
| CN105718184A (en) | Data processing method and apparatus | |
| CN107222526B (en) | Method, device and equipment for pushing promotion information and computer storage medium | |
| CN104090888A (en) | Method and device for analyzing user behavior data | |
| CN110532351A (en) | Recommend word methods of exhibiting, device, equipment and computer readable storage medium | |
| CN111597446A (en) | Content pushing method and device based on artificial intelligence, server and storage medium | |
| CN111460221A (en) | Comment information processing method and device and electronic equipment | |
| CN103038769A (en) | System and method for directing content to users of a social networking engine | |
| CN111815375B (en) | User portrayal method and device in advertisement putting | |
| CN109241455B (en) | Recommended object display method and device | |
| CN110428277B (en) | Touch method of recommended product, storage medium and program product | |
| CN113271325B (en) | Information push method, device, electronic device and computer readable medium | |
| CN109558528A (en) | Article method for pushing, device, computer readable storage medium and server | |
| US20240371162A1 (en) | Computerized system and method for fine-grained event detection and content hosting therefrom | |
| CN105706081A (en) | Structured informational link annotations |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | 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 |