CN105427129A - Information delivery method and system - Google Patents
Information delivery method and system Download PDFInfo
- Publication number
- CN105427129A CN105427129A CN201510768448.2A CN201510768448A CN105427129A CN 105427129 A CN105427129 A CN 105427129A CN 201510768448 A CN201510768448 A CN 201510768448A CN 105427129 A CN105427129 A CN 105427129A
- Authority
- CN
- China
- Prior art keywords
- crowd
- seed crowd
- candidate seed
- information
- weight value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0251—Targeted advertisements
- G06Q30/0269—Targeted advertisements based on user profile or attribute
- G06Q30/0271—Personalized advertisement
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Entrepreneurship & Innovation (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Economics (AREA)
- Game Theory and Decision Science (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses an information delivery method and system. The method comprises the following steps: acquiring a seed crowd, and determining a characteristic weight value vector of the seed crowd through a first preset model; extending the seed crowd based on the characteristic weight value vector to generate candidate seed crowds; according to historical data of information delivery related to the candidate seed crowds, determining the optimal candidate seed crowd; determining a characteristic weight value vector of the optimal candidate seed crowd through a second preset model; and performing secondary extending based on the characteristic weight value vector of the optimal candidate seed crowd to generate a target crowd. According to the method, the optimal candidate seed crowd is determined by combining the historical data of information delivery related to the candidate seed crowds, and then secondary model training and secondary extending are performed to determine the target crowd to which information is delivered; and not-prime seed users who have redundancy or are low in confidence level are filtered, so that the data calculation amount is reduced, the determination of the target crowd can be more accurate, and the advertisement delivery effect is improved.
Description
Technical field
The invention belongs to communication technical field, particularly relate to a kind of put-on method and system of information.
Background technology
Along with the development of Internet technology, various instant messaging and social application are also arisen at the historic moment like the mushrooms after rain.The a large number of users data that instant messaging and social activity are involved in applying, the hobby of such as user, age and demand etc., for the input of information, the input as advertisement has great meaning.
In prior art, by what collect under specific transactions scene, crowd product, service to same requirements and interest is referred to as seed crowd, and the usual quantity of seed crowd is few, generally below 100,000; The crowd with seed crowd with same characteristic features is referred to as expansion crowd, and the quantity of expansion crowd is generally the several times of seed crowd.General, when carrying out advertisement putting, first find expansion crowd by seed crowd, thereafter using the targeted customer of expansion crowd as advertisement putting, when there being multiple seed crowd, first can find the expansion crowd of each seed crowd, then can get the targeted customer of common factor as final advertisement putting of the expansion crowd of each seed crowd.
To in the research and practice process of prior art, the present inventor finds, owing to have chosen the targeted customer of common factor as advertisement putting of the expansion crowd of all seed crowds in prior art, when seed crowd is more, data calculated amount can be caused comparatively huge; Further, redundancy or the non-prime seed user such as with a low credibility may be had in seed crowd, the determination of targeted customer can be impacted, cause the problem that advertisement delivery effect is not good.
Summary of the invention
The object of the present invention is to provide a kind of put-on method and system of information, be intended to reduce data calculated amount, and improve advertisement delivery effect.
For solving the problems of the technologies described above, the embodiment of the present invention provides following technical scheme:
A put-on method for information, comprising:
Obtain seed crowd, and determine the feature weight value vector of described seed crowd by the first preset model;
Based on described feature weight value vector, described seed crowd is expanded, generate corresponding candidate seed crowd;
Add up the historical data that described candidate seed crowd throws in about information, and according to described historical data, determine best candidate seed crowd, described historical data comprises clicking rate;
The feature weight value vector of described best candidate seed crowd is determined by the second preset model;
Based on the feature weight value vector of described best candidate seed crowd, described best candidate seed crowd is expanded, generates target group, to carry out information input to described target group.
For solving the problems of the technologies described above, the embodiment of the present invention also provides following technical scheme:
A jettison system for information, comprising:
Acquiring unit, for obtaining seed crowd;
First determining unit, for the feature weight value vector by the first preset model determination seed crowd;
Expanding element, for expanding described seed crowd based on described feature weight value vector, generates corresponding candidate seed crowd;
Optimum crowd's determining unit, for adding up the historical data that described candidate seed crowd throws in about information, and according to described historical data, determine best candidate seed crowd, described historical data comprises clicking rate;
Second determining unit, for determining the feature weight value vector of described best candidate seed crowd by the second preset model;
Throw in unit, for the feature weight value vector based on described best candidate seed crowd, described best candidate seed crowd is expanded, generates target group, to carry out information input to described target group.
Relative to prior art, the embodiment of the present invention, first by the feature weight value vector of model training determination seed crowd, thereafter feature based weighted value vector is expanded seed crowd, generate candidate seed crowd, and determine best candidate seed crowd in conjunction with candidate seed crowd about the historical data that information is thrown in, secondary expansion is carried out for best candidate seed crowd, the target group that information of determining is thrown in, have redundancy or the non-prime seed user such as with a low credibility that may exist are filtered, not only greatly reduce data calculated amount, and the determination of target group can be made more accurate, improve the input effect of advertisement.
Accompanying drawing explanation
Below in conjunction with accompanying drawing, by the specific embodiment of the present invention describe in detail, will make technical scheme of the present invention and other beneficial effect apparent.
Fig. 1 a is the application scenarios schematic diagram of the jettison system of the information that the embodiment of the present invention provides;
Fig. 1 b is the schematic flow sheet of the put-on method of the information that first embodiment of the invention provides;
The schematic flow sheet of the put-on method of the information that Fig. 2 provides for second embodiment of the invention;
The structural representation of the jettison system of the information that Fig. 3 a provides for third embodiment of the invention;
Another structural representation of the jettison system of the information that Fig. 3 b provides for third embodiment of the invention.
Embodiment
Please refer to graphic, wherein identical element numbers represents identical assembly, and principle of the present invention implements to illustrate in a suitable computing environment.The following description is based on the illustrated specific embodiment of the invention, and it should not be regarded as limiting the present invention not at other specific embodiment that this describes in detail.
In the following description, specific embodiments of the invention illustrate, unless otherwise stating clearly with reference to the step performed by or multi-section computing machine and symbol.Therefore, these steps and operation will have to mention for several times and performed by computing machine, and computing machine execution as referred to herein includes by representing with the operation of the computer processing unit of the electronic signal of the data in a structuring pattern.These data of this operation transformation or the position maintained in the memory system of this computing machine, its reconfigurable or other running changing this computing machine in the mode known by the tester of this area.The data structure that these data maintain is the provider location of this internal memory, and it has the particular characteristics defined by this data layout.But the principle of the invention illustrates with above-mentioned word, it is not represented as a kind of restriction, and this area tester can recognize that the plurality of step of the following stated and operation also may be implemented in the middle of hardware.
Term as used herein " module " can regard the software object as performing in this arithmetic system as.Different assembly as herein described, module, engine and service can be regarded as the objective for implementation in this arithmetic system.And system and method as herein described is preferably implemented in the mode of software, certainly also can implement on hardware, all within scope.
The embodiment of the present invention provides a kind of put-on method and system of information.
See Fig. 1 a, the application scenarios schematic diagram of the jettison system of the information that this figure provides for the embodiment of the present invention, in this application scenarios, comprises information jettison system, be mainly used in obtaining seed crowd, as many groups have the crowd of same requirements and interest to product, service; According to described seed crowd, and vectorial by the feature weight value of the first preset model determination seed crowd; Feature based weighted value vector is expanded seed crowd, generates corresponding candidate seed crowd; The historical data that statistics candidate seed crowd throws in about information, as clicking rate and/or Transaction Information; According to historical data, target group is determined from candidate seed crowd, can be concrete, according to historical data, determine best candidate seed crowd, the feature weight value vector of described best candidate seed crowd is determined by the second preset model, based on the feature weight value vector of described best candidate seed crowd, described best candidate seed crowd is expanded, determines target group, to carry out information input to target group, as carried out advertisement putting etc.
In addition, in this application scenarios, can also database be comprised, be mainly used in storing the seed crowd that magnanimity treats candidate; Certainly, the related data of described seed crowd can derive from information release platform and transaction platform, namely information release platform and transaction platform can also be comprised in this application scenarios, wherein, information release platform be mainly used in record Internet video broadcasting time, point praise rate, point do not like the clicking rate related datas such as rate, and according to the target group that information jettison system is determined, carry out information input, as carried out advertisement putting etc.Transaction platform, is mainly used for the related data of recorded information transaction, as the transaction record of the relation between information broadcasting time and transaction value, user, etc.
To be described in detail respectively below.
First embodiment
In the present embodiment, the angle from information jettison system be described, this information jettison system specifically can be integrated in the network equipments such as server.
A put-on method for information, comprising: obtain seed crowd, and vectorial by the feature weight value of the first preset model determination seed crowd; Feature based weighted value vector is expanded seed crowd, generates corresponding candidate seed crowd; The historical data that statistics candidate seed crowd throws in about information, and according to historical data, determine best candidate seed crowd, historical data comprises clicking rate; By the feature weight value vector of the second preset model determination best candidate seed crowd; Based on the feature weight value vector of best candidate seed crowd, best candidate seed crowd is expanded, generates target group, to carry out information input to target group.
Refer to Fig. 1 b, Fig. 1 b is the schematic flow sheet of the put-on method of the information that first embodiment of the invention provides.Described method comprises:
In step S101, obtain seed crowd, and vectorial by the feature weight value of the first preset model determination seed crowd.
Be understandable that, in the embodiment of the present invention, seed crowd refers to and collects under specific transactions scene, and product, service are had to the crowd of same requirements and interest, seed crowd usual quantity is few, generally below 100,000.
Another it is contemplated that seed crowd can collect specific in presetting database, wherein this database data can be uploaded by the information spinner of various information (as advertisement), also can obtain by the transaction platform of correspondence.
After getting seed crowd, need to analyze this seed crowd, to determine the feature weight value vector of seed crowd, such as, can specifically be comprised by the feature weight value vector of the first preset model determination seed crowd:
(1) seed crowd is defined as the first positive example sample;
(2) obtain non-seed crowd, and described non-seed crowd is defined as the first negative routine sample;
Be understandable that, want to find out the customer group similar to seed crowd from deep bid user, this problem is converted into classical two classification (0,1) problems, namely Output rusults only has two kinds, such as: (male/female), (ill/not have disease), (spam/non-spam email), (enemy army/non-enemy army) etc.
Wherein, positive example sample is called by the sample of labeling in the sample data that in two disaggregated models, training pattern is used, seed crowd belongs to the positive example sample of label under line, be not called negative routine sample by the sample of labeling in the sample data that in two disaggregated models, training pattern is used, the inside but contains potential exemplar.
Such as, suppose that this seed crowd is to the interested user of certain brand panel computer, at this moment seed user bag is defined as the positive sample of disaggregated model training study, be called the first positive example sample herein, from deep bid user, namely in aforementioned presetting database, magnanimity treats candidate seed crowd, in find out corresponding data negative sample (being called the first negative routine sample herein) for model training study.
Further, such as, after deep bid user can being rejected whole seed crowd union, random sampling goes out how group is defined as the first negative routine sample with the crowd of the first positive example sample size equalization.
(3) proper vector of the first positive example sample and the proper vector of the first negative routine sample is determined respectively;
(4) proper vector of the first positive example sample and the first negative routine sampling feature vectors are imported the first preset model and carry out training study, generate the first logistic regression disaggregated model;
(5) corresponding feature weight value vector is determined according to the first logistic regression disaggregated model.
In the embodiment of the present invention, the first positive example sample and the first negative routine sample are exactly to allow two desired values of model learning, interested in or lose interest in certain brand panel computer.
Thus, first preset model can be understood as a disaggregated model, interested for deep bid user or estimation standardization of loseing interest in will be arrived [0 by the purposes of disaggregated model exactly, 1], the probability simultaneously estimated according to this can select a threshold value that classification results is mapped to 0 or 1, so just achieves deep bid user to the prediction interested or uninterested of certain brand panel computer.
That is, import after the first preset model carries out training study, i.e. the exportable feature weight value vector the first positive example sample and the first negative routine sample to discrimination.
Be understandable that, in the embodiment of the present invention, proper vector refers to the characteristic set of the sample of users cut out from the global characteristics of deep bid user, and global characteristics refers to the summation a certain deep bid user being chosen the user interest classification in multiple social platform, such as, the global characteristics that Tengxun's deep bid user is corresponding specifically can comprise QQ group's categorize interests classification, micro-letter public number categorize interests classification, advertisement business categorize interests classification, mobile phone A PP series etc.
Preferably, for the selection of disaggregated model, in this embodiment, employ logistic regression (LogisticRegression, LR) model that is comparatively ripe, versatility better, high latitude expansion sex excellence; It is contemplated that, select the model of other logics such as decision tree (DecisionTree, DT), support vector machine (SupportVectorMachine, SVM) that classification can be realized equally, be only herein and illustrate, do not form limitation of the invention.
In step s 102, feature based weighted value vector is expanded seed crowd, generates corresponding candidate seed crowd.
Such as, seed crowd expansion can be specific as follows:
1, expand described seed crowd according to feature weight value vector, be expanded crowd;
2, candidate seed crowd is defined as by meeting pre-conditioned expansion crowd.
Wherein, expansion crowd refers to the crowd with seed crowd with same characteristic features, and its quantity is generally the several times of seed crowd.
After the crowd of being expanded, candidate seed crowd can be defined as by meeting pre-conditioned expansion crowd, such as, the candidate user corresponding to the answer interval [0.5 ~ 1] of two classification problems be defined as candidate seed crowd.
In step s 103, the historical data that statistics candidate seed crowd throws in about information, and according to historical data, determine best candidate seed crowd.
In embodiments of the present invention, this historical data can specifically comprise clicking rate and/or Transaction Information;
Such as, adding up candidate seed crowd can be specific as follows about the historical data that information is thrown in:
A, acquisition presupposed information release platform and/or the database corresponding to default transaction platform;
B, in the database obtained, add up the historical data that candidate seed crowd throws in about information.
Be understandable that, historical data comprised clicking rate and/or Transaction Information, and wherein clicking rate can obtain in a certain preset time period, as one week, one month etc.; Mainly can comprise rate of a little praising (the some number of praising/exposure number), put and do not like rate (point does not like number/exposure number), comment rate (filling in comment number/exposure number) etc.Transaction Information mainly can comprise the information such as number of transaction, transaction value of product or service.
Be understandable that, determine that the mode of best candidate seed crowd has a lot, such as: according to historical data, to candidate seed, crowd sorts, and based on sequence, the candidate seed crowd meeting preset rules is defined as best candidate seed crowd.
In step S104, determined the feature weight value vector of described best candidate seed crowd by the second preset model.
In step S105, based on the feature weight value vector of best candidate seed crowd, best candidate seed crowd is expanded, generates target group, to carry out information input to target group.
Wherein, described step S104 and step S105 can be specially:
Based on best candidate seed crowd, carry out second training model and assessment and secondary crowd and expand, and then determine target group, such as, the determination of the feature weight value vector of best candidate seed crowd can specifically comprise:
A, this best candidate seed crowd is defined as the second positive example sample;
B, from aforementioned candidate seed crowd, determine the second negative routine sample;
C, determine the proper vector of the second positive example sample and the proper vector of the second negative routine sample respectively;
D, the proper vector of the second positive example sample and the second negative routine sampling feature vectors are imported the second preset model carry out training study, generate the second logistic regression disaggregated model;
E, to determine corresponding feature weight value vector according to this second logistic regression disaggregated model.
Be understandable that, at second training model with assessment, the process that the feature weight value vector of best candidate seed crowd is determined can with reference to the vectorial process determined of the feature weight value of the above Ziren group, in secondary crowd expansion, the process that best candidate seed crowd expands specifically with reference to the aforementioned process expanded seed crowd, can be repeated no more herein.
From the above, the put-on method of the information that the present embodiment provides, first by the feature weight value vector of model training determination seed crowd, thereafter feature based weighted value vector is expanded seed crowd, generate candidate seed crowd, and determine best candidate seed crowd in conjunction with candidate seed crowd about the historical data that information is thrown in, secondary expansion is carried out for best candidate seed crowd, the target group that information of determining is thrown in, have redundancy or the non-prime seed user such as with a low credibility that may exist are filtered, not only greatly reduce data calculated amount, and the determination of target group can be made more accurate, improve the input effect of advertisement.
Second embodiment
According to the method described by the first embodiment, below citing is described in further detail.
Refer to Fig. 2, the schematic flow sheet of the put-on method of the information that Fig. 2 provides for second embodiment of the invention.The embodiment of the present invention to propose based on expansion crowd in the wide clicking rate of passing network, sells value, point is praised, put foundations such as not liking, as the important references index of assessment seed crowd quality, thus in seed crowd, select the method that optimum seed and its extending user carry out advertisement putting.The method comprises:
In step s 201, information jettison system obtains many group seed crowds.
First, collect many group seed crowds from different channel, such as, can be CRM user data under line, also can be the line label data of each product data team, etc.
Can be concrete, such as, seed crowd can collect in presetting database, and wherein this database data can be uploaded by the information spinner of various information (as advertisement), also can obtain by the transaction platform of correspondence, not do concrete restriction herein.
In step S202, information jettison system determines the first positive example sample and the first negative routine sample.
In the embodiment of the present invention, suppose that seed crowd is to the interested user of certain brand panel computer, now, aforementioned many group seed crowds are defined as the positive sample of disaggregated model training study, for ease of distinguishing and setting forth, this positive sample is called the first positive example sample herein, accordingly, after deep bid user being rejected whole seed crowd union, random sampling goes out how group is defined as the first negative routine sample with the crowd of seed crowd positive example sample size equalization.
In step S203, information jettison system, by model training, determines the feature weight value vector of many group seed crowds.
Such as, the proper vector organizing the first positive example sample and the first negative routine sample is imported the first preset model more and carries out training study, create multiple logistic regression LR disaggregated model, namely exportable many groups to the first positive and negative routine sample have discrimination feature weight value vector, wherein, the feature with discrimination refers to the feature the first positive example sample and the first negative routine sample can being carried out distinguishing.As, if the feature of the first positive example sample listens song for liking, the feature of the first negative routine sample for not like listening song, then, listens the feature of song for having discrimination.
Be understandable that, in the embodiment of the present invention, proper vector refers to the characteristic set of the sample of users cut out from the global characteristics of deep bid user, and global characteristics refers to the summation a certain deep bid user being chosen the user interest classification in multiple social platform, such as, the global characteristics that Tengxun's deep bid user is corresponding specifically can comprise QQ group's categorize interests classification, micro-letter public number categorize interests classification, advertisement business categorize interests classification, mobile phone A PP series etc.
Such as, can be concrete, by seed crowd characteristic vector from cutting global characteristics out, preferably, this feature can not need hand picking, entirely selects in global characteristics; And non-seed crowd is reject seed crowd union in deep bid user after, random sampling goes out the crowd of many groups and seed crowd positive example sample size equalization, again from cutting proper vector global characteristics, now seed crowd, non-seed crowd characteristic vector are ready.
In step S204, information jettison system carries out crowd's expansion to seed crowd, generates many group candidate seed crowds.
According to the feature weight value vector that step S203 training result exports, calculate the global characteristics of deep bid user respectively, obtain many group expansion crowds, the mode wherein expanded has a lot, such as can expand according to each eigenwert, or principal character value is carried out expanding etc.
Further, be that the user of more than 0.5 is considered as candidate user by prediction probability, thus export many group candidate seed crowds, that is, the candidate user corresponding to the answer interval [0.5 ~ 1] of two classification problems is defined as candidate seed crowd.
In step S205, the historical data that information jettison system statistics candidate seed crowd throws in about information.
Wherein, this historical data comprises clicking rate and/or Transaction Information;
Such as, many groups candidate seed crowd that step S205 exports is calculated historical overall clicking rate respectively and sells value, or clicking rate under certain release platform, information of the same type, sell values, point praises rate, put and do not like rate.
In step S206, information jettison system determination best candidate seed crowd.
With reference to the history advertisement effectiveness data corresponding to above-mentioned each group of candidate seed crowd, according to different demand can by clicking rate, sell be worth carry out integrated ordered to all candidate crowds of expanding, user forward for rank is chosen out, as best candidate seed crowd.
In step S207, information jettison system is trained by secondary model, determines the feature weight value vector of best candidate seed crowd.
Specifically comprise:
(1) the second positive example sample and the second negative routine sample is determined;
(2) proper vector of the second positive example sample and the proper vector of the second negative routine sample is determined respectively;
(3) proper vector of the second positive example sample and the second negative routine sampling feature vectors are imported the second preset model and carry out training study, generate the second logistic regression disaggregated model;
(4) corresponding feature weight value vector is determined according to this second logistic regression disaggregated model.
The best candidate seed crowd determined step S206 is as the second positive example sample, and after rejecting best candidate seed crowd in candidate seed crowd, random sampling goes out one group and is defined as the second negative routine sample with the crowd of the second positive example sample size equalization.
Such as, can be concrete, by best candidate seed crowd characteristic vector from cutting global characteristics out, preferably, this feature can not need hand picking, entirely selects in global characteristics; Again by the non-optimal candidate seed crowd that determines from cutting proper vector global characteristics, now best candidate seed crowd, non-optimal candidate seed crowd characteristic vector are ready, import the second preset model, namely disaggregated model carries out training study, namely exportable one group to the second positive and negative routine sample have discrimination feature weight value vector.
Preferably, for the selection of disaggregated model, in this embodiment, employ logistic regression LR model that is comparatively ripe, versatility better, high latitude expansion sex excellence; It is contemplated that select the model of other logics can realize classification equally as decision tree DT, support vector machines, be only herein and illustrate, do not form limitation of the invention.
In step S208, information jettison system carries out secondary crowd expansion, generates target group.
According to many stack features weight that previous step training result exports, the global characteristics of deep bid user is calculated, obtain the expansion crowd of best candidate seed crowd, further, by prediction probability be more than 0.5 extending user be considered as final information and throw in target group, that is, the extending user corresponding to the answer interval [0.5 ~ 1] of two classification problems is defined as target group.
Be understandable that, in some more accurately scene, can also carry out third time model training and third time crowd expansion, repeat no more herein.
From the above, the put-on method of the information that the present embodiment provides, first by the feature weight value vector of model training determination seed crowd, thereafter feature based weighted value vector is expanded seed crowd, generate candidate seed crowd, and determine target group in conjunction with the historical data that candidate seed crowd throws in about information, have redundancy or the non-prime seed user such as with a low credibility that may exist is filtered; That is, the present invention shows according to the historical information effect of extending user, comprise the important references index as assessment ratio for input and output such as clicking rate, transaction value, not only greatly reduce data calculated amount, and the determination of target group can be made more accurate, improve the input effect of advertisement; Further, by secondary model training study and secondary crowd expansion, determine best candidate seed crowd, thus determine optimum target group, such as, based on preceding method, the advertisement putting targeted customer of final delineation is expanded by optimum seed crowd, can ensure that choosing of targeted customer is optimal result, can realize carrying out advertisement putting accurately to target group.
3rd embodiment
For ease of better implementing the put-on method of the information that the embodiment of the present invention provides, the embodiment of the present invention also provides a kind of system of the put-on method based on above-mentioned information.Wherein the implication of noun is identical with the method for the input of above-mentioned information, and specific implementation details can explanation in reference method embodiment.
Refer to Fig. 3 a, the structural representation of the jettison system of the information that Fig. 3 a provides for the embodiment of the present invention, the jettison system of described information can comprise acquiring unit 301, first determining unit 302, expanding element 303, optimum crowd's determining unit 304, second determining unit 305 and throw in unit 306.
Wherein, acquiring unit 301, for obtaining seed crowd; Determining unit 302, for according to described seed crowd, and by the feature weight value vector of the first preset model determination seed crowd.
Be understandable that, in the embodiment of the present invention, seed crowd refers to and collects under specific transactions scene, and product, service are had to the crowd of same requirements and interest, seed crowd usual quantity is few, generally below 100,000.
Another it is contemplated that seed crowd can collect specific in presetting database, wherein this database data can be uploaded by the information spinner of various information (as advertisement), also can obtain by the transaction platform of correspondence.
Expanding element 303, for expanding described seed crowd based on described feature weight value vector, generates corresponding candidate seed crowd; Optimum crowd's determining unit 304, for adding up the historical data that described candidate seed crowd throws in about information, and according to described historical data, determine best candidate seed crowd, described historical data comprises clicking rate.
Thereafter, second determining unit 305, for being determined the feature weight value vector of described best candidate seed crowd by the second preset model, throw in unit 306, for the feature weight value vector based on described best candidate seed crowd, described best candidate seed crowd is expanded, generates target group, to carry out information input to described target group.
Can with reference to figure 3b, another structural representation of the jettison system of the information provided for the embodiment of the present invention, described first determining unit 302 can specifically comprise:
(1) first determines subelement 3021, for described seed crowd is defined as the first positive example sample;
(2) second determine subelement 3022, for obtaining non-seed crowd, and described non-seed crowd are defined as the first negative routine sample;
Such as, suppose that this seed crowd is to the interested user of certain brand panel computer, at this moment seed user bag is defined as the positive sample of disaggregated model training study, be called the first positive example sample herein, from deep bid user, namely in aforementioned presetting database, magnanimity treats candidate seed crowd, in find out corresponding data negative sample (being called the first negative routine sample herein) for model training study.
Further, such as, after deep bid user can being rejected whole seed crowd union, random sampling goes out how group is defined as the first negative routine sample with the crowd of the first positive example sample size equalization.
(3) the 3rd determine subelement 3023, for the proper vector of the negative routine sample of proper vector and first of determining the first positive example sample respectively;
(4) first model generation subelements 3024, carry out training study for the proper vector of the first positive example sample and the first negative routine sampling feature vectors are imported the first preset model, generate the first logistic regression disaggregated model;
(5) the 4th determine subelement 3025, for determining corresponding feature weight value vector according to the first logistic regression disaggregated model.
Be understandable that, want to find out the customer group similar to seed crowd from deep bid user, this problem is converted into classical two classification (0,1) problems, namely Output rusults only has two kinds, such as: (male/female) (ill/not have disease) (spam/non-spam email) (enemy army/non-enemy army).
In the embodiment of the present invention, the first positive example sample and the first negative routine sample are exactly to allow two desired values of model learning, interested in or lose interest in certain brand panel computer.
Thus, first preset model can be understood as a disaggregated model, interested for deep bid user or estimation standardization of loseing interest in will be arrived [0 by the purposes of disaggregated model exactly, 1], the probability simultaneously estimated according to this can select a threshold value that classification results is mapped to 0 or 1, so just achieves deep bid user to the prediction interested or uninterested of certain brand panel computer.
That is, import after the first preset model carries out training study, i.e. the exportable feature weight value vector the first positive example sample and the first negative routine sample to discrimination.
Be understandable that, in the embodiment of the present invention, proper vector refers to the characteristic set of the sample of users cut out from the global characteristics of deep bid user, and global characteristics refers to the summation a certain deep bid user being chosen the user interest classification in multiple social platform, such as, the global characteristics that Tengxun's deep bid user is corresponding specifically can comprise QQ group's categorize interests classification, micro-letter public number categorize interests classification, advertisement business categorize interests classification, mobile phone A PP series etc.
Preferably, for the selection of disaggregated model, in this embodiment, employ logistic regression LR model that is comparatively ripe, versatility better, high latitude expansion sex excellence; It is contemplated that select the model of other logics can realize classification equally as decision tree DT, support vector machines, be only herein and illustrate, do not form limitation of the invention.
Based on this, described expanding element 303 can be specifically for:
Expand described seed crowd according to described feature weight value vector, be expanded crowd; Candidate seed crowd is defined as by meeting pre-conditioned expansion crowd.
Wherein, expansion crowd refers to the crowd with seed crowd with same characteristic features, and its quantity is generally the several times of seed crowd.
After the crowd of being expanded, candidate seed crowd can be defined as by meeting pre-conditioned expansion crowd, such as, the candidate user corresponding to the answer interval [0.5 ~ 1] of two classification problems be defined as candidate seed crowd.
Further, described optimum crowd's determining unit 304 adds up the historical data that candidate seed crowd throws in about information, can be specifically for:
Obtain presupposed information release platform and/or the database corresponding to default transaction platform, add up the historical data that described candidate seed crowd throws in about information in a database.
Be understandable that, historical data comprised clicking rate and/or Transaction Information, and wherein clicking rate can obtain in a certain preset time period, as one week, one month etc.; Mainly can comprise rate of a little praising (the some number of praising/exposure number), put and do not like rate (point does not like number/exposure number), comment rate (filling in comment number/exposure number) etc.Transaction Information mainly can comprise the information such as number of transaction, transaction value of product or service.
Be understandable that, determine that the mode of best candidate seed crowd has a lot, preferably, optimum crowd's determining unit 304, specifically for according to described historical data, can also sort to described candidate seed crowd; Based on described sequence, the candidate seed crowd meeting preset rules is defined as best candidate seed crowd.
Based on best candidate seed crowd, carry out second training model and assessment and secondary crowd and expand, and then determine target group, such as, described second determining unit 305 determines the feature weight value vector of best candidate seed crowd, can specifically comprise:
5th determines subelement 3051, for described best candidate seed crowd is defined as the second positive example sample;
6th determines subelement 3052, for determining the second negative routine sample from described candidate seed crowd;
7th determines subelement 3053, for the proper vector of the proper vector and described second negative routine sample of determining described second positive example sample respectively;
Second model generation subelement 3054, carries out training study for the proper vector of described second positive example sample and described second negative routine sampling feature vectors are imported the second preset model, generates the second logistic regression disaggregated model;
8th determines subelement 3055, for determining corresponding feature weight value vector according to described second logistic regression disaggregated model.
Be understandable that, at second training model with assessment, the process that the feature weight value vector of best candidate seed crowd is determined can with reference to the vectorial process determined of the feature weight value of the above Ziren group, in secondary crowd expansion, the process that best candidate seed crowd expands specifically with reference to the aforementioned process expanded seed crowd, can be repeated no more herein.
During concrete enforcement, above unit can realize as independently entity, and can carry out combination in any yet, realize as same or several entities, the concrete enforcement of above unit see embodiment of the method above, can not repeat them here.
The jettison system of this information specifically can be integrated in the network equipments such as server.
From the above, the jettison system of the information that the present embodiment provides, first by the feature weight value vector of model training determination seed crowd, thereafter feature based weighted value vector is expanded seed crowd, generate candidate seed crowd, and determine best candidate seed crowd in conjunction with candidate seed crowd about the historical data that information is thrown in, secondary expansion is carried out for best candidate seed crowd, the target group that information of determining is thrown in, have redundancy or the non-prime seed user such as with a low credibility that may exist are filtered, not only greatly reduce data calculated amount, and the determination of target group can be made more accurate, improve the input effect of advertisement.
In the above-described embodiments, the description of each embodiment is all emphasized particularly on different fields, there is no the part described in detail in certain embodiment, see above for the detailed description of the put-on method of information, can repeat no more herein.
The jettison system of the described information that the embodiment of the present invention provides, be for example computing machine, panel computer, the mobile phone with touch function etc., the put-on method of the information in the jettison system of described information and foregoing embodiments belongs to same design, the jettison system of described information can be run the either method provided in the put-on method embodiment of described information, its specific implementation process refers to the put-on method embodiment of described information, repeats no more herein.
It should be noted that, for the put-on method of information of the present invention, this area common test personnel are appreciated that all or part of flow process of the put-on method realizing information described in the embodiment of the present invention, that the hardware that can control to be correlated with by computer program has come, described computer program can be stored in a computer read/write memory medium, as being stored in the storer of terminal, and performed by least one processor in this terminal, can comprise in the process of implementation as described in the flow process of embodiment of put-on method of information.Wherein, described storage medium can be magnetic disc, CD, ROM (read-only memory) (ROM, ReadOnlyMemory), random access memory (RAM, RandomAccessMemory) etc.
For the jettison system of the described information of the embodiment of the present invention, its each functional module can be integrated in a process chip, also can be that the independent physics of modules exists, also can two or more module integrations in a module.Above-mentioned integrated module both can adopt the form of hardware to realize, and the form of software function module also can be adopted to realize.If described integrated module using the form of software function module realize and as independently production marketing or use time, also can be stored in a computer read/write memory medium, described storage medium such as be ROM (read-only memory), disk or CD etc.
Above the put-on method of a kind of information that the embodiment of the present invention provides and system are described in detail, apply specific case herein to set forth principle of the present invention and embodiment, the explanation of above embodiment just understands method of the present invention and core concept thereof for helping; Meanwhile, for those skilled in the art, according to thought of the present invention, all will change in specific embodiments and applications, in sum, this description should not be construed as limitation of the present invention.
Claims (12)
1. a put-on method for information, is characterized in that, comprising:
Obtain seed crowd, and determine the feature weight value vector of described seed crowd by the first preset model;
Based on described feature weight value vector, described seed crowd is expanded, generate corresponding candidate seed crowd;
Add up the historical data that described candidate seed crowd throws in about information, and according to described historical data, determine best candidate seed crowd, described historical data comprises clicking rate;
The feature weight value vector of described best candidate seed crowd is determined by the second preset model;
Based on the feature weight value vector of described best candidate seed crowd, described best candidate seed crowd is expanded, generates target group, to carry out information input to described target group.
2. the put-on method of information according to claim 1, is characterized in that, the described feature weight value vector being determined described seed crowd by the first preset model, being comprised:
Described seed crowd is defined as the first positive example sample;
Obtain non-seed crowd, and described non-seed crowd is defined as the first negative routine sample;
Determine the proper vector of described first positive example sample and the proper vector of described first negative routine sample respectively;
The proper vector of described first positive example sample and described first negative routine sampling feature vectors are imported the first preset model and carries out training study, generate the first logistic regression disaggregated model;
Corresponding feature weight value vector is determined according to described first logistic regression disaggregated model.
3. the put-on method of information according to claim 2, is characterized in that, describedly expands described seed crowd based on described feature weight value vector, generates corresponding candidate seed crowd, comprising:
Expand described seed crowd according to described feature weight value vector, be expanded crowd;
Candidate seed crowd is defined as by meeting pre-conditioned expansion crowd.
4. the put-on method of the information according to any one of claims 1 to 3, is characterized in that, the historical data that the described candidate seed crowd of described statistics throws in about information, comprising:
Obtain presupposed information release platform and/or the database corresponding to default transaction platform;
Add up the historical data that described candidate seed crowd throws in about information in the database.
5. the put-on method of information according to claim 4, is characterized in that, described according to described historical data, determines best candidate seed crowd, comprising:
According to described historical data, described candidate seed crowd is sorted;
Based on described sequence, the candidate seed crowd meeting preset rules is defined as best candidate seed crowd.
6. the put-on method of information according to claim 5, is characterized in that, the described feature weight value vector passing through the second preset model determination best candidate seed crowd, comprising:
Described best candidate seed crowd is defined as the second positive example sample;
The second negative routine sample is determined from described candidate seed crowd;
Determine the proper vector of described second positive example sample and the proper vector of described second negative routine sample respectively;
The proper vector of described second positive example sample and described second negative routine sampling feature vectors are imported the second preset model and carries out training study, generate the second logistic regression disaggregated model;
Corresponding feature weight value vector is determined according to described second logistic regression disaggregated model.
7. a jettison system for information, is characterized in that, comprising:
Acquiring unit, for obtaining seed crowd;
First determining unit, for the feature weight value vector by the first preset model determination seed crowd;
Expanding element, for expanding described seed crowd based on described feature weight value vector, generates corresponding candidate seed crowd;
Optimum crowd's determining unit, for adding up the historical data that described candidate seed crowd throws in about information, and according to described historical data, determine best candidate seed crowd, described historical data comprises clicking rate;
Second determining unit, for determining the feature weight value vector of described best candidate seed crowd by the second preset model;
Throw in unit, for the feature weight value vector based on described best candidate seed crowd, described best candidate seed crowd is expanded, generates target group, to carry out information input to described target group.
8. the jettison system of information according to claim 7, is characterized in that, described first determining unit comprises:
First determines subelement, for described seed crowd is defined as the first positive example sample;
Second determines subelement, for obtaining non-seed crowd, and described non-seed crowd is defined as the first negative routine sample;
3rd determines subelement, for the proper vector of the proper vector and described first negative routine sample of determining described first positive example sample respectively;
First model generation subelement, carries out training study for the proper vector of described first positive example sample and described first negative routine sampling feature vectors are imported the first preset model, generates the first logistic regression disaggregated model;
4th determines subelement, for determining corresponding feature weight value vector according to described first logistic regression disaggregated model.
9. the jettison system of information according to claim 8, is characterized in that, described expanding element specifically for:
Expand described seed crowd according to described feature weight value vector, be expanded crowd; Candidate seed crowd is defined as by meeting pre-conditioned expansion crowd.
10. the jettison system of the information according to any one of claim 7 to 9, is characterized in that, described optimum crowd determine specifically for:
Obtain presupposed information release platform and/or the database corresponding to default transaction platform, add up the historical data that described candidate seed crowd throws in about information in the database.
The jettison system of 11. information according to claim 10, is characterized in that, described optimum crowd's determining unit, specifically for according to described historical data, sorts to described candidate seed crowd; Based on described sequence, the candidate seed crowd meeting preset rules is defined as best candidate seed crowd.
The jettison system of 12. information according to claim 11, is characterized in that, described second determining unit comprises:
5th determines subelement, for described best candidate seed crowd is defined as the second positive example sample;
6th determines subelement, for determining the second negative routine sample from described candidate seed crowd;
7th determines subelement, for the proper vector of the proper vector and described second negative routine sample of determining described second positive example sample respectively;
Second model generation subelement, carries out training study for the proper vector of described second positive example sample and described second negative routine sampling feature vectors are imported the second preset model, generates the second logistic regression disaggregated model;
8th determines subelement, for determining corresponding feature weight value vector according to described second logistic regression disaggregated model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510768448.2A CN105427129B (en) | 2015-11-12 | 2015-11-12 | Information delivery method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510768448.2A CN105427129B (en) | 2015-11-12 | 2015-11-12 | Information delivery method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105427129A true CN105427129A (en) | 2016-03-23 |
CN105427129B CN105427129B (en) | 2020-09-04 |
Family
ID=55505314
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510768448.2A Active CN105427129B (en) | 2015-11-12 | 2015-11-12 | Information delivery method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105427129B (en) |
Cited By (27)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105931079A (en) * | 2016-04-29 | 2016-09-07 | 合网络技术(北京)有限公司 | Method and apparatus for diffusing seed users |
CN107633421A (en) * | 2017-08-24 | 2018-01-26 | 武汉科技大学 | A kind of processing method and processing device of market prediction data |
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 |
CN107729937A (en) * | 2017-10-12 | 2018-02-23 | 北京京东尚科信息技术有限公司 | For determining the method and device of user interest label |
CN107918922A (en) * | 2017-11-15 | 2018-04-17 | 中国联合网络通信集团有限公司 | Business recommended method and business recommended device |
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 |
WO2018126953A1 (en) * | 2017-01-06 | 2018-07-12 | 腾讯科技(深圳)有限公司 | Seed population expanding method, device, information releasing system and storing medium |
CN108427690A (en) * | 2017-02-15 | 2018-08-21 | 腾讯科技(深圳)有限公司 | Information distribution method and device |
CN109034896A (en) * | 2018-07-23 | 2018-12-18 | 北京奇艺世纪科技有限公司 | Crowd's prediction technique and device are launched in a kind of advertisement |
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 |
CN109359236A (en) * | 2018-08-22 | 2019-02-19 | 中国平安人寿保险股份有限公司 | Target user's method for digging and device based on big data analysis |
CN109801091A (en) * | 2017-11-16 | 2019-05-24 | 腾讯科技(深圳)有限公司 | Targeted user population localization method, device, computer equipment and storage medium |
CN109961310A (en) * | 2017-12-26 | 2019-07-02 | 北京奇虎科技有限公司 | Advertisement placement method, server and computer readable storage medium |
CN110059244A (en) * | 2019-02-01 | 2019-07-26 | 阿里巴巴集团控股有限公司 | Audient's extended method and device |
CN110110195A (en) * | 2019-05-07 | 2019-08-09 | 宜人恒业科技发展(北京)有限公司 | A kind of impurity sweep-out method and device |
WO2020043001A1 (en) * | 2018-08-27 | 2020-03-05 | 阿里巴巴集团控股有限公司 | Advertisement placement method, method for determining popularization population, server, and terminal |
CN111178934A (en) * | 2019-11-29 | 2020-05-19 | 北京深演智能科技股份有限公司 | Method and device for acquiring target object |
CN111460299A (en) * | 2020-03-31 | 2020-07-28 | 北京深演智能科技股份有限公司 | Information delivery method and device |
CN111563761A (en) * | 2020-01-19 | 2020-08-21 | 深圳前海微众银行股份有限公司 | Crowd expansion method, device, equipment and storage medium |
CN111582944A (en) * | 2020-05-14 | 2020-08-25 | 北京金山安全软件有限公司 | Crowd expansion method, device, equipment and storage medium for advertisement |
CN111882360A (en) * | 2020-07-30 | 2020-11-03 | 北京达佳互联信息技术有限公司 | User group expansion method and device |
CN111881364A (en) * | 2020-07-22 | 2020-11-03 | 北京深演智能科技股份有限公司 | Media file delivery method and device, storage medium and processor |
CN112053184A (en) * | 2020-08-20 | 2020-12-08 | 腾讯科技(深圳)有限公司 | Promotion information delivery method and device, electronic equipment and storage medium |
US10990500B2 (en) | 2018-05-18 | 2021-04-27 | Beijing Didi Infinity Technology And Development Co., Ltd. | Systems and methods for user analysis |
CN112950276A (en) * | 2021-03-12 | 2021-06-11 | 华扬联众数字技术(深圳)有限公司 | Seed population expansion method based on multi-order feature combination |
CN112967100A (en) * | 2021-04-02 | 2021-06-15 | 杭州网易云音乐科技有限公司 | Similar population expansion method, device, computing equipment and medium |
CN113011922A (en) * | 2021-03-18 | 2021-06-22 | 北京百度网讯科技有限公司 | Similar population determination method and device, electronic equipment and storage medium |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101071439A (en) * | 2007-05-24 | 2007-11-14 | 北京交通大学 | Interactive video searching method based on multi-view angle |
US20080140589A1 (en) * | 2006-12-06 | 2008-06-12 | Microsoft Corporation | Active learning framework for automatic field extraction from network traffic |
CN102456199A (en) * | 2010-10-18 | 2012-05-16 | 北京学之途网络科技有限公司 | Method and device for expanding internet user sample set and acquiring attribute parameter |
CN103105924A (en) * | 2011-11-15 | 2013-05-15 | 中国科学院深圳先进技术研究院 | Man-machine interaction method and device |
US20130124439A1 (en) * | 2010-05-24 | 2013-05-16 | Nec Corporation | Information extraction system, method, and program |
CN103136267A (en) * | 2011-12-01 | 2013-06-05 | 腾讯科技(深圳)有限公司 | Expert mining method and device based on community |
CN103150578A (en) * | 2013-04-09 | 2013-06-12 | 山东师范大学 | Training method of SVM (Support Vector Machine) classifier based on semi-supervised learning |
CN104537383A (en) * | 2015-01-20 | 2015-04-22 | 全国组织机构代码管理中心 | Massive organizational structure data classification method and system based on particle swarm |
CN104751354A (en) * | 2015-04-13 | 2015-07-01 | 合一信息技术(北京)有限公司 | Advertisement cluster screening method |
CN104834641A (en) * | 2014-02-11 | 2015-08-12 | 腾讯科技(北京)有限公司 | Processing method of network medium information and relevant system |
CN105023165A (en) * | 2014-04-17 | 2015-11-04 | 腾讯科技(深圳)有限公司 | Method, device and system for controlling release tasks in social networking platform |
-
2015
- 2015-11-12 CN CN201510768448.2A patent/CN105427129B/en active Active
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080140589A1 (en) * | 2006-12-06 | 2008-06-12 | Microsoft Corporation | Active learning framework for automatic field extraction from network traffic |
CN101071439A (en) * | 2007-05-24 | 2007-11-14 | 北京交通大学 | Interactive video searching method based on multi-view angle |
US20130124439A1 (en) * | 2010-05-24 | 2013-05-16 | Nec Corporation | Information extraction system, method, and program |
CN102456199A (en) * | 2010-10-18 | 2012-05-16 | 北京学之途网络科技有限公司 | Method and device for expanding internet user sample set and acquiring attribute parameter |
CN103105924A (en) * | 2011-11-15 | 2013-05-15 | 中国科学院深圳先进技术研究院 | Man-machine interaction method and device |
CN103136267A (en) * | 2011-12-01 | 2013-06-05 | 腾讯科技(深圳)有限公司 | Expert mining method and device based on community |
CN103150578A (en) * | 2013-04-09 | 2013-06-12 | 山东师范大学 | Training method of SVM (Support Vector Machine) classifier based on semi-supervised learning |
CN104834641A (en) * | 2014-02-11 | 2015-08-12 | 腾讯科技(北京)有限公司 | Processing method of network medium information and relevant system |
CN105023165A (en) * | 2014-04-17 | 2015-11-04 | 腾讯科技(深圳)有限公司 | Method, device and system for controlling release tasks in social networking platform |
CN104537383A (en) * | 2015-01-20 | 2015-04-22 | 全国组织机构代码管理中心 | Massive organizational structure data classification method and system based on particle swarm |
CN104751354A (en) * | 2015-04-13 | 2015-07-01 | 合一信息技术(北京)有限公司 | Advertisement cluster screening method |
Cited By (38)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105931079A (en) * | 2016-04-29 | 2016-09-07 | 合网络技术(北京)有限公司 | Method and apparatus for diffusing seed users |
US11301884B2 (en) | 2017-01-06 | 2022-04-12 | Tencent Technology (Shenzhen) Company Limited | Seed population diffusion method, device, information delivery system and storage medium |
WO2018126953A1 (en) * | 2017-01-06 | 2018-07-12 | 腾讯科技(深圳)有限公司 | Seed population expanding method, device, information releasing system and storing medium |
CN108427690A (en) * | 2017-02-15 | 2018-08-21 | 腾讯科技(深圳)有限公司 | Information distribution method and device |
CN107633421A (en) * | 2017-08-24 | 2018-01-26 | 武汉科技大学 | A kind of processing method and processing device of market prediction data |
CN107729937A (en) * | 2017-10-12 | 2018-02-23 | 北京京东尚科信息技术有限公司 | For determining the method and device of user interest label |
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 |
CN107918922A (en) * | 2017-11-15 | 2018-04-17 | 中国联合网络通信集团有限公司 | Business recommended method and business recommended device |
CN107918922B (en) * | 2017-11-15 | 2020-10-27 | 中国联合网络通信集团有限公司 | Service recommendation method and service recommendation device |
CN109801091B (en) * | 2017-11-16 | 2022-12-20 | 腾讯科技(深圳)有限公司 | Target user group positioning method and device, computer equipment and storage medium |
CN109801091A (en) * | 2017-11-16 | 2019-05-24 | 腾讯科技(深圳)有限公司 | Targeted user population localization method, device, computer equipment and storage medium |
CN109961310A (en) * | 2017-12-26 | 2019-07-02 | 北京奇虎科技有限公司 | Advertisement placement method, server and computer readable storage medium |
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 |
US10990500B2 (en) | 2018-05-18 | 2021-04-27 | Beijing Didi Infinity Technology And Development Co., Ltd. | Systems and methods for user analysis |
CN109034896A (en) * | 2018-07-23 | 2018-12-18 | 北京奇艺世纪科技有限公司 | Crowd's prediction technique and device are launched in a kind of advertisement |
CN109359236A (en) * | 2018-08-22 | 2019-02-19 | 中国平安人寿保险股份有限公司 | Target user's method for digging and device based on big data analysis |
WO2020043001A1 (en) * | 2018-08-27 | 2020-03-05 | 阿里巴巴集团控股有限公司 | Advertisement placement method, method for determining popularization population, server, and terminal |
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 |
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 |
CN110059244A (en) * | 2019-02-01 | 2019-07-26 | 阿里巴巴集团控股有限公司 | Audient's extended method and device |
CN110110195A (en) * | 2019-05-07 | 2019-08-09 | 宜人恒业科技发展(北京)有限公司 | A kind of impurity sweep-out method and device |
CN111178934A (en) * | 2019-11-29 | 2020-05-19 | 北京深演智能科技股份有限公司 | Method and device for acquiring target object |
CN111178934B (en) * | 2019-11-29 | 2024-03-08 | 北京深演智能科技股份有限公司 | Method and device for acquiring target object |
CN111563761A (en) * | 2020-01-19 | 2020-08-21 | 深圳前海微众银行股份有限公司 | Crowd expansion method, device, equipment and storage medium |
CN111563761B (en) * | 2020-01-19 | 2024-06-07 | 深圳前海微众银行股份有限公司 | Crowd expansion method, device, equipment and storage medium |
CN111460299A (en) * | 2020-03-31 | 2020-07-28 | 北京深演智能科技股份有限公司 | Information delivery method and device |
CN111582944A (en) * | 2020-05-14 | 2020-08-25 | 北京金山安全软件有限公司 | Crowd expansion method, device, equipment and storage medium for advertisement |
CN111582944B (en) * | 2020-05-14 | 2023-08-22 | 北京金山安全软件有限公司 | Crowd expansion method, device and equipment for advertisement and storage medium |
CN111881364A (en) * | 2020-07-22 | 2020-11-03 | 北京深演智能科技股份有限公司 | Media file delivery method and device, storage medium and processor |
CN111882360A (en) * | 2020-07-30 | 2020-11-03 | 北京达佳互联信息技术有限公司 | User group expansion method and device |
CN112053184A (en) * | 2020-08-20 | 2020-12-08 | 腾讯科技(深圳)有限公司 | Promotion information delivery method and device, electronic equipment and storage medium |
CN112053184B (en) * | 2020-08-20 | 2024-01-30 | 腾讯科技(深圳)有限公司 | Popularization information delivery method and device, electronic equipment and storage medium |
CN112950276B (en) * | 2021-03-12 | 2022-03-29 | 华扬联众数字技术(深圳)有限公司 | Seed population expansion method based on multi-order feature combination |
CN112950276A (en) * | 2021-03-12 | 2021-06-11 | 华扬联众数字技术(深圳)有限公司 | Seed population expansion method based on multi-order feature combination |
CN113011922B (en) * | 2021-03-18 | 2023-08-04 | 北京百度网讯科技有限公司 | Method and device for determining similar crowd, electronic equipment and storage medium |
CN113011922A (en) * | 2021-03-18 | 2021-06-22 | 北京百度网讯科技有限公司 | Similar population determination method and device, electronic equipment and storage medium |
CN112967100A (en) * | 2021-04-02 | 2021-06-15 | 杭州网易云音乐科技有限公司 | Similar population expansion method, device, computing equipment and medium |
CN112967100B (en) * | 2021-04-02 | 2024-03-15 | 杭州网易云音乐科技有限公司 | Similar crowd expansion method, device, computing equipment and medium |
Also Published As
Publication number | Publication date |
---|---|
CN105427129B (en) | 2020-09-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105427129A (en) | Information delivery method and system | |
WO2020147594A1 (en) | Method, system, and device for obtaining expression of relationship between entities, and advertisement retrieval system | |
CN108205766A (en) | Information-pushing method, apparatus and system | |
CN105023165A (en) | Method, device and system for controlling release tasks in social networking platform | |
CN108563680A (en) | Resource recommendation method and device | |
CN111523831B (en) | Risk group identification method and device, storage medium and computer equipment | |
CN109933666A (en) | A kind of good friend's automatic classification method, device, computer equipment and storage medium | |
CN105022754A (en) | Social network based object classification method and apparatus | |
CN113435509B (en) | Small sample scene classification and identification method and system based on meta-learning | |
WO2021208535A1 (en) | Recommendation method and device based on automatic feature grouping | |
CN101685456B (en) | Search method, system and device | |
CN103646070A (en) | Data processing method and device for search engine | |
US20210073669A1 (en) | Generating training data for machine-learning models | |
CN105320764A (en) | 3D model retrieval method and 3D model retrieval apparatus based on slow increment features | |
CN112733035A (en) | Knowledge point recommendation method and device based on knowledge graph, storage medium and electronic device | |
CN103700004A (en) | Method and device for pushing microblog advertising service information | |
Moniz et al. | A framework for recommendation of highly popular news lacking social feedback | |
CN102866997A (en) | Method and device for processing user data | |
CN114049161B (en) | E-commerce big data feedback-based push optimization method and E-commerce big data system | |
CN106709572A (en) | Data processing method and equipment | |
CN110069695A (en) | Label processing method and device | |
CN112733036A (en) | Knowledge point recommendation method and device, storage medium and electronic device | |
CN113259495A (en) | Block chain-based news fixed-point delivery method and system and readable storage medium | |
CN111882360A (en) | User group expansion method and device | |
Xiao et al. | Cost-sensitive semi-supervised ensemble model for customer churn prediction |
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 |