CN108985804A - Flow stage division and device - Google Patents
Flow stage division and device Download PDFInfo
- Publication number
- CN108985804A CN108985804A CN201710398014.7A CN201710398014A CN108985804A CN 108985804 A CN108985804 A CN 108985804A CN 201710398014 A CN201710398014 A CN 201710398014A CN 108985804 A CN108985804 A CN 108985804A
- Authority
- CN
- China
- Prior art keywords
- information
- flow
- model
- traffic
- data
- 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.)
- Pending
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
- G06Q30/0277—Online advertisement
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
Landscapes
- Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Engineering & Computer Science (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Entrepreneurship & Innovation (AREA)
- Economics (AREA)
- Game Theory and Decision Science (AREA)
- Marketing (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention provides a kind of flow stage division and device, this method comprises: receiving the flow information that the SDK acquisition device in each client acquires the mobile Internet ad traffic of upload respectively；The data factors for assessing classification are determined according to each flow information, quality grading collective model are constructed by machine learning, and the composite quality index of each flow information is generated according to the quality grading collective model；Each flow information is classified according to each composite quality index.The present invention determines the data factors assessed for mobile terminal discharge characteristic using the data on flows of mobile terminal, then constructs quality grading collective model by machine learning, and final realize accurately is classified mobile Internet ad traffic.
Description
Technical field
This application involves mobile internet technical fields, and in particular to a kind of flow stage division and device.
Background technique
With gradually popularizing for mobile device, mobile Internet in fast-developing trend, the online application of mobile interchange,
The flows such as advertisement also increase in succession, and center of gravity is transferred in mobile Internet by more and more company and products, while movement is set
The standby time used and frequency have also been more than non-mobile device gradually.
Advertisement is a kind of realization mode of flow current main-stream, and the quality of flow directly affects it and cashes efficiency, advertiser
It tends to obtain higher input-output ratio on high-quality flow, advertiser is more willing to carry out advertisement throwing on high-quality flow
It puts, thus high-quality flow often just has higher cashability.Conversely, low-quality flow is enough because advertiser can not be obtained
Budget favors its cashability with regard to relatively poor, at the same time, appears in the advertisement on low-quality flow, builds to the brand of advertiser
If may also bring a negative impact.Therefore, the assessment and classification of flow mass cashes flow and risk control just seems non-
It is often valuable.
There are following both sides defects for existing flow stage division: on the one hand, existing flow stage division is main
For the internet traffic at non-moving end, it is not directed to mobile terminal Internet advertising flow, mobile terminal is not made full use of to interconnect
Distinctive data carry out assessment classification in net advertisement；On the other hand, it is comprehensive to lack a standard for existing flow stage division
It is unified and can in evaluation process Optimized model to improve the assessment models of accuracy rate, to mobile Internet ad traffic into
Row assessment classification.
Summary of the invention
In view of drawbacks described above in the prior art or deficiency, it is intended to provide a kind of utilization distinctive data in mobile terminal, building
And the comprehensive unified assessment models of optimisation criteria, thus the flow classification side being accurately classified to mobile Internet ad traffic
Method and device.
In a first aspect, the present invention provides a kind of flow stage division, this method comprises:
Receive the flow letter that the SDK acquisition device in each client acquires the mobile Internet ad traffic of upload respectively
Breath；
The data factors for assessing classification are determined according to each flow information, and it is comprehensive to construct quality grading by machine learning
Model, and generate according to the quality grading collective model composite quality index of each ad traffic；
Each ad traffic is classified according to each composite quality index.
Second aspect, the present invention provide another flow stage division, this method comprises:
Acquire the flow information of mobile Internet ad traffic；
The flow information is uploaded, so that server-side determines the data factors for assessing classification according to the flow information, is led to
Machine learning building quality grading collective model is crossed, is referred to according to the comprehensive quality that the quality grading collective model generates ad traffic
Number, is classified ad traffic according to the composite quality index.
The third aspect, the present invention provide a kind of flow grading plant, the device include communication unit, model construction unit and
Stage unit.
Wherein, the SDK acquisition device that communication unit is configured to receive in each client acquires the movement of upload respectively
The flow information of Internet advertising flow；Model construction unit is configured to be determined according to each flow information for assessing classification
Data factors construct quality grading collective model by machine learning, and generate each advertisement according to the quality grading collective model
The composite quality index of flow；Stage unit is configured to be classified each ad traffic according to each composite quality index.
Fourth aspect, the present invention also provides a kind of equipment, including one or more processors and memory, wherein memory
Comprising can by instruction that the one or more processors execute so that the one or more processors execute it is each according to the present invention
The flow stage division that embodiment provides.
5th aspect, the present invention also provides a kind of computer readable storage medium for being stored with computer program, the calculating
Machine program makes computer execute the flow stage division that each embodiment provides according to the present invention.
The flow stage division and device that many embodiments of the present invention provide using the data on flows of mobile terminal by being determined
Quality grading collective model is constructed for the data factors that mobile terminal discharge characteristic is assessed, then by machine learning, finally
Realization is accurately classified mobile Internet ad traffic；
The flow stage division and device that some embodiments of the invention provide further pass through building single factor test quality grading
Model denoises data set, to optimize the data for being used for training quality hierarchical synthesis model, obtains more accurately classification mould
Type, to keep final classification results more accurate；
The flow stage division and device that some embodiments of the invention provide are further by tying training result and classification
Fruit carries out double verification, and the accuracy of classification results has been ensured while advanced optimizing model.
Detailed description of the invention
By reading a detailed description of non-restrictive embodiments in the light of the attached drawings below, the application's is other
Feature, objects and advantages will become more apparent upon:
Fig. 1 is a kind of flow chart for flow stage division that one embodiment of the invention provides.
Fig. 2 is a kind of flow chart of preferred embodiment of method shown in Fig. 1.
Fig. 3 is the flow chart of step S13 in a kind of preferred embodiment of method shown in Fig. 1.
Fig. 4 is the flow chart of step S133 in a kind of preferred embodiment of method shown in Fig. 3.
Fig. 5 is the flow chart of step S135 in a kind of preferred embodiment of method shown in Fig. 3.
Fig. 6 is the flow chart of step S15 in a kind of preferred embodiment of method shown in Fig. 1.
Fig. 7 is the flow chart for another flow stage division that one embodiment of the invention provides.
Fig. 8 is a kind of structural schematic diagram for flow grading plant that one embodiment of the invention provides.
Fig. 9 is a kind of structural schematic diagram of preferred embodiment of Fig. 8 shown device.
Figure 10 is the structural schematic diagram of model construction unit in a kind of preferred embodiment of Fig. 8 shown device.
Figure 11 is a kind of preferred structure schematic diagram of model construction unit shown in Figure 10.
Figure 12 is the structural schematic diagram of stage unit in a kind of preferred embodiment of Fig. 8 shown device.
Figure 13 is a kind of structural schematic diagram for equipment that one embodiment of the invention provides.
Specific embodiment
The application is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is used only for explaining related invention, rather than the restriction to the invention.It also should be noted that in order to
Convenient for description, part relevant to invention is illustrated only in attached drawing.
It should be noted that in the absence of conflict, the features in the embodiments and the embodiments of the present application can phase
Mutually combination.The application is described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.
Fig. 1 is a kind of flow chart for flow stage division that one embodiment of the invention provides.
As shown in Figure 1, in the present embodiment, the present invention provides a kind of flow stage division suitable for server-side, the party
Method includes:
S11: the stream that the SDK acquisition device in each client acquires the mobile Internet ad traffic of upload respectively is received
Measure information；
S13: the data factors for assessing classification are determined according to each flow information, quality grading is constructed by machine learning
Collective model, and generate according to the quality grading collective model composite quality index of each ad traffic；
S15: each ad traffic is classified according to each composite quality index.
In the present embodiment, SDK acquisition device is to be embedded in the application program for being installed on client, for (in user
Under conditions of permission) flow information of each ad traffic of acquisition and the Software Development Kit (Software that is uploaded
Development Kit, abbreviation SDK), client is the mobile devices such as mobile phone, plate.Flow information specifically includes following three classes
Information: source-information, advertising data information and user behavior information.
Wherein, source-information includes host application information, source country information and source device information；Ad data
Information includes that the mount message of application, advertisement are recommended in advertisement exposure information, ad click information, advertisement download information, advertisement
Take in information and ad margins information；User behavior information includes the unloading information that application is recommended in advertisement, and insertion SDK data are adopted
The retention information of the application program of acquisition means.
In more embodiments, can according to actual needs it configure the classification of flow information to any one in above-mentioned three classes
Class or multiclass, and can further comprise other different classes of flow informations, it is as long as can be uploaded by the acquisition of SDK acquisition device
It can；Equally can according to actual needs respectively by the Specific disposition of source-information, advertising data information, user behavior information be it is above-mentioned
It is any one or more in every terms of information, and can further comprise other different information.For example, in the present embodiment, weight
Point concern user recommends advertisement the unloading behavior of application, and in another embodiment, user behavior information may also include, advertisement
Institute's recommendation recommendation behavioural information (such as user whether by the content forward circle of friends, whether by the application such as microblogging into
Row recommend, etc.) etc. different types of information.
In step s 11, as long as being equipped with the application program for being embedded in the SDK acquisition device in the mobile device of user, i.e.,
Can be acquired by the SDK acquisition device the flow information of each mobile Internet ad traffic and uploaded in a mobile device.
Server-side receives the flow information that the SDK acquisition device of each mobile device uploads respectively.For example, in the mobile device of user's first
SDK acquisition device has detected that advertisement A, B, C generate ad traffic a, b, c respectively, acquires the flow of ad traffic a, b, c respectively
Information is simultaneously uploaded to server-side；SDK acquisition device has detected that advertisement A, B, D generate advertisement respectively in the mobile device of user's second
Flow a, b, d acquire the flow information of ad traffic a, b, d respectively and are uploaded to server-side；Deng.
In step s 13, server-side according to the received each flow information of step S11 determine for assess the data being classified because
Element.
In the present embodiment, data factors include: unloading rate, retention ratio, clicking rate, installation rate and every thousand exposure advertisements
Take in Ecpm.
Specifically, unloading rate is determined according to the installation and discharging quantity that count mount message and unloading information acquisition respectively:
Unloading rate=discharging quantity/installation (1)
After user has viewed the popularization advertisement of some application program, interested user can download installation, and this applies journey
Sequence；And when user using this application program feel without what value when, be just likely to uninstall this application program.
Therefore, unloading rate can directly reflect that the fancy grade for the application program that user promotes advertisement, unloading rate height show user
Low to the degree of recognition of this application program, the quality of the ad traffic is relatively poor.
Retention ratio retains the next day retention amount of information acquisition according to statistics and newly-increased retention amount determines:
Retention ratio=next day retention amount/newly-increased retention amount (2)
The height of retention ratio reflects user to the frequency of use and the degree of recognition of this product, thus to a certain extent
Also can reflect the quality of the ad traffic: retention ratio height illustrates that the application program payes attention to user experience, will not allow easily wide
Bad user is cracked using the experience in product, the flow side of the advertisement more focuses on advertisement compared to the flow side of general advertisement
The presentation skill of accurate push, the creative design of advertisement, advertisement.Therefore, retention ratio height can usually reflect the ad traffic
Quality is relatively preferable.
Clicking rate is determined according to the click volume and light exposure that count ad click information and advertisement exposure information acquisition respectively:
Clicking rate=click volume/light exposure (3)
Installation rate is determined according to installation and click volume:
Installation rate=installation/click volume (4)
Clicking rate and installation rate can react ad traffic respectively and click, from the conversion feelings for clicking installation from being exposed to
Condition.But while it is noted that the ad traffic of clicking rate and installation rate exception, there are certain cheating risk, advertisements
Flow Value is not equally high.Therefore, during the present embodiment subsequent builds quality grading collective model, clicking rate and installation
It is lower that the weight of rate compares other three item datas factors.
Every thousand exposure advertising income Ecpm are determined according to the ad margins and light exposure of statistics ad margins information acquisition:
Ecpm=(ad margins * 1000)/light exposure (5)
Ecpm can react the situation of Profit of flow.The higher ad traffic bring income of Ecpm is higher；Ecpm is lower
Ad traffic bring income it is lower.
In more embodiments, by data factors above-mentioned five can be configured by the practical information category for including according to flow information
It is any one or more in, and can further comprise the data factors of the quality good or not of other reaction ad traffics, as long as
It is obtained using ad traffic data in mobile terminal included in flow information, and can reflect the quality good or not of ad traffic, i.e.,
Same technique effect can be achieved.
After determining all data factor, machine learning model commonly used in the art, structure are combined according to all data factor
Quality grading collective model is built, the flow information of each ad traffic is inputted into the quality grading collective model, generates each ad stream
The composite quality index of amount.
In step S15, by the way that the critical value for determining and being classified is trained and is rule of thumb finely adjusted to model, thus
Each ad traffic is classified according to the composite quality index that critical value and step S13 are generated.
Above-described embodiment is by determining the number assessed for mobile terminal discharge characteristic using the data on flows of mobile terminal
Quality grading collective model is constructed according to factor, then by machine learning, final realize carries out standard to mobile Internet ad traffic
Really classification.
Fig. 2 is a kind of flow chart of preferred embodiment of method shown in Fig. 1.
As shown in Fig. 2, in a preferred embodiment, after step S11 further include:
S12: it is integrated according to the classification that source-information carries out different dimensions to each flow information.Wherein, dimension is according to source
The type of information determines.
Specifically, in the present embodiment, since source-information includes host application information, source country information and is come
Source device information, therefore, dimension include application program dimension, national dimension and equipment dimension.It, can basis in more embodiments
Different source-informations configures different types of dimension, and customized other dimensions can be further configured.
By being combined to above-mentioned every dimension, can be provided in subsequent data screening and model construction process wide
Accuse data information of the flow under separate sources dimension granularity.
Fig. 3 is the flow chart of step S13 in a kind of preferred embodiment of method shown in Fig. 1.
As shown in figure 3, in a preferred embodiment, step S13 includes:
S131: the data set for constructing model is filtered out from each flow information；
S132: the data factors for assessing classification are determined according to each flow information；
S133: single factor test quality grading model is constructed according to the data factors, and according to the single factor test quality grading
Model obtains the single factor test performance figure of each ad traffic；
S134: the data set is denoised according to the single factor test performance figure；
S135: according to the data set after denoising, constructing quality grading collective model by machine learning and statistical method,
To obtain the composite quality index of each ad traffic.
Specifically, it in step S131, can be screened by way of following at least one:
The history ranked data for inquiring each ad traffic, marks whether to be classified according to history ranked data；
Mark whether to be classified according to the advertising data information of the flow information after classification integration；
Rule of thumb strategy is screened.
In step S134, the single factor test performance figure obtained for step S133 is very prominent or obvious unreasonable wide
Flow is accused, can take and screen out data, be directly classified, the denoisings such as manual examination and verification, to ensure that it will not influence machine
The training of learning model and the building of quality grading collective model.
Fig. 4 is the flow chart of step S133 in a kind of preferred embodiment of method shown in Fig. 3.
As shown in figure 4, in a preferred embodiment, step S133 is specifically included:
S1331: the single datum factor of each ad traffic is calculated, and calculated result is ranked up；
S1332: the high-quality distributed area and non-prime distributed area of single datum factor are determined according to ranking results；
S1333: the desired value of single datum factor is determined according to high-quality distributed area and non-prime distributed area；
S1334: single factor test quality grading model is constructed according to desired value and logistic regression formula；
S1335: the single factor test performance figure of each ad traffic is obtained according to single factor test quality grading model.
Specifically, by taking unloading rate as an example, in step S1331, the unloading of each ad traffic in data set is calculated separately out
Rate, and be ranked up from small to large；
In step S1332, the data set after being ranked up by unloading rate is traversed, the unloading rate for marking high-quality flow is found out
Distributed area and the unloading rate distributed area for marking non-prime flow, as high-quality distributed area and non-prime distributed area；
In step S1333, a desired value of the expectation unloading rate as unloading rate is set, and by adjusting making the phase
Prestige value is between high-quality distributed area and non-prime distributed area；
In step S1334, according to the step S1333 desired value determined and logistic regression formula building single factor test quality point
Grade model:
Wherein, VUnloading rateIt (i) is the unloading rate performance figure of ad traffic, i is the serial number of ad traffic, UEXPFor unloading rate
Desired value, UiFor the unloading rate of ad traffic；
In step S1335, by the single factor test quality grading mould of the unloading rate input step S1334 generation of each ad traffic
Type obtains the unloading rate performance figure of each ad traffic.
In more embodiments, different formula building single factor test quality grading models also can be used, if can generate from
Scattered, reflection ad traffic quality single factor test performance figure is, it can be achieved that identical technical effect.
Above-described embodiment further passes through building single factor test quality grading model and denoises to data set, is used for optimization
The data of training quality hierarchical synthesis model, obtain more accurate hierarchy model, to keep final classification results more accurate.
Fig. 5 is the flow chart of step S135 in a kind of preferred embodiment of method shown in Fig. 3.
As shown in figure 5, in a preferred embodiment, step S135 includes:
S1351: the data set input machine learning model after denoising is trained, to training result combination statistics side
Method is to construct generation quality grading collective model and obtain classification results；
S1352: classification results are verified according to corresponding history ranked data and/or artificial labeled data:
If authentication failed, S1353 is thened follow the steps: the model structure in adjustment machine learning model and/or statistical method
Parameter is built, and return step S1351 rebuilds quality grading collective model；
If being proved to be successful, S1354 is thened follow the steps: being generated according to quality grading collective model and logistic regression formula each wide
Accuse the composite quality index of flow.
Fig. 6 is the flow chart of step S15 in a kind of preferred embodiment of method shown in Fig. 1.
As shown in fig. 6, in a preferred embodiment, step S15 includes:
S151: it determines composite quality index critical value, and each ad traffic is classified and is marked；
S153: verifying classification results according to corresponding history ranked data, if it exists notable difference then mark with
For manual examination and verification；
S155: classification results are shown.
Above-described embodiment further by carrying out double verification to training result and classification results, advanced optimizes model
The accuracy of classification results has been ensured simultaneously.
Fig. 7 is the flow chart for another flow stage division that one embodiment of the invention provides.Method shown in Fig. 7 can adopt
Collection uploads the flow information of ad traffic to cooperate method shown in FIG. 1.
As shown in fig. 7, in the present embodiment, the present invention also provides another flow classifications for being suitable for SDK acquisition device
Method, comprising:
S21: the flow information of acquisition mobile Internet ad traffic；
S23: uploading the flow information, for server-side according to the flow information determine for assess be classified data because
Element constructs quality grading collective model by machine learning, the synthesis of ad traffic is generated according to the quality grading collective model
Performance figure is classified ad traffic according to the composite quality index.
The information category that flow information specifically includes method shown in Figure 1, details are not described herein again.
Fig. 8 is a kind of structural schematic diagram for flow grading plant that one embodiment of the invention provides.Device shown in Fig. 8 can
It is corresponding to execute method shown in FIG. 1.
As shown in figure 8, in the present embodiment, the present invention provides a kind of flow grading plant 10, including communication unit 11, mould
Type construction unit 13 and stage unit 15.
Wherein, communication unit 11, which is configured to receive the SDK acquisition device in each client 20 and acquires respectively, uploads
The flow information of mobile Internet ad traffic；
Model construction unit 13 is configured to determine the data factors for assessing classification according to each flow information, passes through machine
Device study building quality grading collective model, and referred to according to the comprehensive quality that the quality grading collective model generates each ad traffic
Number；
Stage unit 15 is configured to be classified each ad traffic according to each composite quality index.
The model construction and flow classification principle of the flow grading plant 10 method shown in Figure 1, it is no longer superfluous herein
It states.
Fig. 9 is a kind of structural schematic diagram of preferred embodiment of Fig. 8 shown device.Device shown in Fig. 9 can be corresponded to and be held
Row method shown in Fig. 2.
As shown in figure 9, in a preferred embodiment, flow grading plant 10 further includes classification integral unit 12.
Classification integral unit 12 is configured to be integrated according to the classification that source-information carries out different dimensions to each flow information.
Concrete principle method shown in Figure 2, details are not described herein again.
Figure 10 is the structural schematic diagram of model construction unit in a kind of preferred embodiment of Fig. 8 shown device.Figure 10 institute
Showing device can correspond to method shown in execution Fig. 3-4.
As shown in Figure 10, in a preferred embodiment, model construction unit 13 includes data set screening subelement 131, number
Subelement 132, unifactor model building subelement 133, denoising subelement 134 and composite factor model construction are determined according to factor
Unit 135.
Data set screening subelement 131 is configured to filter out the data set for constructing model from each flow information；
Data factors determine subelement 132 be configured to according to each flow information determine for assess be classified data because
Element；
Unifactor model building subelement 133 is configured to construct single factor test quality grading model according to data factors, and
The single factor test performance figure of each ad traffic is obtained according to single factor test quality grading model；
Denoising subelement 134 is configured to denoise data set according to single factor test performance figure；
Composite factor model construction subelement 135 is configured to pass through machine learning and system according to the data set after denoising
It counts method and constructs quality grading collective model, to obtain the composite quality index of each ad traffic.
In a preferred embodiment, unifactor model building subelement 133 is configurable to execute every step shown in Fig. 4
Suddenly.
Figure 11 is a kind of preferred structure schematic diagram of model construction unit shown in Figure 10.Device shown in Figure 11 can be corresponded to and be held
Row method shown in fig. 5.
As shown in figure 11, in a preferred embodiment, model construction unit 13 further includes the first verifying subelement 136.
First verifying subelement 136 is configured to verify the classification results of quality hierarchical synthesis model: if verifying
Failure, then adjust the model construction parameter in machine learning model and/or statistical method, for composite factor model construction
Unit 135 rebuilds quality grading collective model.
Composite factor model construction subelement 135 is further configured to after the first verifying subelement 136 is proved to be successful
The composite quality index of each ad traffic is generated according to quality grading collective model and logistic regression formula.
Figure 12 is the structural schematic diagram of stage unit in a kind of preferred embodiment of Fig. 8 shown device.Shown in Figure 12
Device, which can correspond to, executes method shown in fig. 6.
As shown in figure 12, in a preferred embodiment, stage unit 15 includes that classification verifying of subelement 151, second is single
Member 153 and result display unit 155.
Wherein, classification subelement 151 is configured to determine composite quality index critical value, and divides each ad traffic
Grade and mark；
Second verifying subelement 153 is configured to verify classification results according to corresponding history ranked data, if
There are notable differences then to mark for manual examination and verification；
As a result display unit 155 is configured to show classification results.
Figure 13 is a kind of structural schematic diagram for equipment that one embodiment of the invention provides.
As shown in figure 13, as on the other hand, present invention also provides a kind of equipment 1300, including one or more centers
Processing unit (CPU) 1301, can be according to the program being stored in read-only memory (ROM) 1302 or from storage section
1308 programs being loaded into random access storage device (RAM) 1303 and execute various movements appropriate and processing.In RAM1303
In, it is also stored with equipment 1300 and operates required various programs and data.CPU1301, ROM1302 and RAM1303 pass through total
Line 1304 is connected with each other.Input/output (I/O) interface 1305 is also connected to bus 1304.
I/O interface 1305 is connected to lower component: the importation 1306 including keyboard, mouse etc.；Including such as cathode
The output par, c 1307 of ray tube (CRT), liquid crystal display (LCD) etc. and loudspeaker etc.；Storage section including hard disk etc.
1308；And the communications portion 1309 of the network interface card including LAN card, modem etc..Communications portion 1309 passes through
Communication process is executed by the network of such as internet.Driver 1310 is also connected to I/O interface 1305 as needed.It is detachable to be situated between
Matter 1311, such as disk, CD, magneto-optic disk, semiconductor memory etc. are mounted on as needed on driver 1310, so as to
In being mounted into storage section 1308 as needed from the computer program read thereon.
Particularly, in accordance with an embodiment of the present disclosure, the flow stage division of any of the above-described embodiment description can be implemented
For computer software programs.For example, embodiment of the disclosure includes a kind of computer program product comprising be tangibly embodied in
Computer program on machine readable media, the computer program include the program code for executing flow stage division.
In such embodiments, which can be downloaded and installed from network by communications portion 1309, and/or from
Detachable media 1311 is mounted.
As another aspect, present invention also provides a kind of computer readable storage medium, the computer-readable storage mediums
Matter can be computer readable storage medium included in the device of above-described embodiment；It is also possible to individualism, it is unassembled
Enter the computer readable storage medium in equipment.Computer-readable recording medium storage has one or more than one program, should
Program is used to execute the flow stage division for being described in the application by one or more than one processor.
Flow chart and block diagram in attached drawing are illustrated according to the system of various embodiments of the invention, method and computer journey
The architecture, function and operation in the cards of sequence product.In this regard, each box in flowchart or block diagram can generation
A part of one module, program segment or code of table, a part of the module, program segment or code include one or more use
The executable instruction of the logic function as defined in realizing.It should also be noted that in some implementations as replacements, being marked in box
The function of note can also occur in a different order than that indicated in the drawings.For example, two boxes succeedingly indicated are actually
It can be basically executed in parallel, they can also be executed in the opposite order sometimes, this is depending on related function.Also it wants
It is noted that the combination of each box in block diagram and or flow chart and the box in block diagram and or flow chart, Ke Yitong
The dedicated hardware based system of functions or operations as defined in executing is crossed to realize, or by specialized hardware and can be calculated
The combination of machine instruction is realized.
Being described in the embodiment of the present application involved unit or module can be realized by way of software, can also be with
It is realized by way of hardware.Described unit or module also can be set in the processor, for example, each unit can
To be the software program being arranged in computer or intelligent movable equipment, it is also possible to the hardware device being separately configured.Wherein, this
The title of a little units or module does not constitute the restriction to the unit or module itself under certain conditions.
Above description is only the preferred embodiment of the application and the explanation to institute's application technology principle.Those skilled in the art
Member is it should be appreciated that invention scope involved in the application, however it is not limited to technology made of the specific combination of above-mentioned technical characteristic
Scheme, while should also cover in the case where not departing from the application design, appointed by above-mentioned technical characteristic or its equivalent feature
Other technical solutions of meaning combination and formation.Such as features described above and (but being not limited to) disclosed herein have similar functions
Technical characteristic replaced mutually and the technical solution that is formed.
Claims (20)
1. a kind of flow stage division characterized by comprising
Receive the flow information that the SDK acquisition device in each client acquires the mobile Internet ad traffic of upload respectively；
The data factors for assessing classification are determined according to each flow information, and it is comprehensive to construct quality grading by machine learning
Model, and generate according to the quality grading collective model composite quality index of each ad traffic；
Each ad traffic is classified according to each composite quality index.
2. flow stage division according to claim 1, which is characterized in that the flow information includes source-information, wide
Accuse data information and user behavior information.
3. flow stage division according to claim 2, which is characterized in that the source-information includes following at least one
: host application information, source country information, source device information；
The advertising data information includes at least one of the following: advertisement exposure information, ad click information, advertisement download information,
The mount message of application, advertising income information, ad margins information are recommended in advertisement；
The user behavior information includes at least one of the following: that the unloading information of application is recommended in advertisement, is embedded in the SDK data
The retention information of the application program of acquisition device.
4. flow stage division according to claim 3, which is characterized in that the data factors include following at least one
: unloading rate, retention ratio, clicking rate, installation rate, every thousand exposure advertising incomes；
The unloading rate is determined according to the installation and discharging quantity that count the mount message and unloading information acquisition respectively；
The retention ratio is determined according to the next day retention amount and newly-increased retention amount that count the retention information acquisition；
The clicking rate is according to the click volume and exposure for counting the ad click information and the advertisement exposure information acquisition respectively
Light quantity determines；
The installation rate is determined according to the installation and the click volume；
Every thousand exposure advertising incomes are according to the ad margins and the light exposure for counting the ad margins information acquisition
It determines.
5. flow stage division according to claim 2, which is characterized in that further include:
It is integrated according to the classification that the source-information carries out different dimensions to each flow information；Wherein, the dimension according to
The type of the source-information determines.
6. flow stage division according to claim 1-5, which is characterized in that described to be believed according to each flow
Breath determines the data factors for assessing classification, constructs quality grading collective model by machine learning, and according to the quality
The composite quality index that hierarchical synthesis model generates each ad traffic includes:
The data set for constructing model is filtered out from each flow information；
The data factors for assessing classification are determined according to each flow information；
Single factor test quality grading model is constructed according to the data factors, and is obtained respectively according to the single factor test quality grading model
The single factor test performance figure of ad traffic；
The data set is denoised according to the single factor test performance figure；
According to the data set after denoising, quality grading collective model is constructed by machine learning and statistical method, to obtain
The composite quality index of each ad traffic.
7. flow stage division according to claim 6, which is characterized in that described to be filtered out from each flow information
Data set for constructing model includes at least one of the following:
The history ranked data for inquiring each ad traffic marks whether point according to the history ranked data
Grade；
Rule of thumb strategy is screened.
8. flow stage division according to claim 6, which is characterized in that described to construct Dan Yin according to the data factors
Quality amount hierarchy model, and obtain according to the single factor test quality grading model single factor test of each ad traffic in the data set
Performance figure includes:
The single datum factor of each ad traffic is calculated, and calculated result is ranked up；
The high-quality distributed area and non-prime distributed area of the single datum factor are determined according to ranking results；
The desired value of the single datum factor is determined according to the high-quality distributed area and non-prime distributed area；
Single factor test quality grading model is constructed according to the desired value and logistic regression formula；
The single factor test performance figure of each ad traffic is obtained according to the single factor test quality grading model.
9. flow stage division according to claim 6, which is characterized in that the data set according to after denoising passes through
Machine learning and statistical method construct quality grading collective model, thus after being denoised in data set each ad traffic it is comprehensive
Closing performance figure includes:
Data set input machine learning model after denoising is trained, to training result combination statistical method to construct life
At quality grading collective model and obtain classification results；
The classification results are verified according to corresponding history ranked data and/or artificial labeled data:
If authentication failed, the model construction parameter in the machine learning model and/or the statistical method is adjusted, is laid equal stress on
New building quality grading collective model；
If being proved to be successful, the synthesis matter of each ad traffic is generated according to the quality grading collective model and logistic regression formula
Volume index.
10. flow stage division according to claim 1-5, which is characterized in that described according to each synthesis
Performance figure carries out classification to each ad traffic
It determines composite quality index critical value, and each ad traffic is classified and is marked；
Classification results are verified according to corresponding history ranked data, notable difference then marks for manually examining if it exists
Core；
Show classification results.
11. a kind of flow stage division characterized by comprising
Acquire the flow information of mobile Internet ad traffic；
The flow information is uploaded, so that server-side determines the data factors for assessing classification according to the flow information, is led to
Machine learning building quality grading collective model is crossed, the synthesis of the ad traffic is generated according to the quality grading collective model
Performance figure is classified the ad traffic according to the composite quality index.
12. flow stage division according to claim 11, which is characterized in that the flow information include source-information,
Advertising data information and user behavior information.
13. flow stage division according to claim 12, which is characterized in that the source-information includes following at least one
: host application information, source country information, source device information；
The advertising data information includes at least one of the following: advertisement exposure information, ad click information, advertisement download information,
The mount message of application, advertising income information, ad margins information are recommended in advertisement；
The user behavior information includes at least one of the following: that the unloading information of application is recommended in advertisement, is embedded in the SDK data
The retention information of the application program of acquisition device.
14. a kind of flow grading plant characterized by comprising
Communication unit, be configured to receive the SDK acquisition device in each client acquire respectively upload mobile Internet it is wide
Accuse the flow information of flow；
Model construction unit is configured to determine the data factors for assessing classification according to each flow information, passes through machine
Device study constructs quality grading collective model, and the synthesis of each ad traffic is generated according to the quality grading collective model
Performance figure；
Stage unit is configured to be classified each ad traffic according to each composite quality index.
15. flow grading plant according to claim 14, which is characterized in that the flow information include source-information,
Advertising data information and user behavior information；
Described device further include:
It is whole to be configured to the classification for carrying out different dimensions to each flow information according to the source-information for classification integral unit
It closes；Wherein, the dimension is determined according to the type of the source-information.
16. flow grading plant according to claim 15, which is characterized in that the model construction unit includes:
Data set screens subelement, is configured to filter out the data set for constructing model from each flow information；
Data factors determine subelement, are configured to determine the data factors for assessing classification according to each flow information；
Unifactor model constructs subelement, is configured to construct single factor test quality grading model, and root according to the data factors
The single factor test performance figure of each ad traffic is obtained according to the single factor test quality grading model；
Subelement is denoised, is configured to denoise the data set according to the single factor test performance figure；
Composite factor model construction subelement is configured to pass through machine learning and statistics side according to the data set after denoising
Method constructs quality grading collective model, to obtain the composite quality index of each ad traffic.
17. flow grading plant according to claim 16, which is characterized in that the model construction unit further include:
First verifying subelement, is configured to verify the classification results of quality hierarchical synthesis model:
If authentication failed, the model construction parameter in machine learning model and/or statistical method is adjusted, for the synthesis
Factor Model building subelement rebuilds quality grading collective model；
The composite factor model construction subelement is further configured to the root after the first verifying subelement is proved to be successful
The composite quality index of each ad traffic is generated according to the quality grading collective model and logistic regression formula.
18. the described in any item flow grading plants of 4-17 according to claim 1, which is characterized in that the stage unit includes:
It is classified subelement, is configured to determine composite quality index critical value, and each ad traffic is classified and is marked；
Second verifying subelement, is configured to verify classification results according to corresponding history ranked data, bright if it exists
The different then mark of significant difference is for manual examination and verification；
As a result display unit is configured to show classification results.
19. a kind of equipment, which is characterized in that the equipment includes:
One or more processors；
Memory, for storing one or more programs,
When one or more of programs are executed by one or more of processors, so that one or more of processors
Execute such as method of any of claims 1-13.
20. a kind of computer readable storage medium for being stored with computer program, which is characterized in that the program is executed by processor
Shi Shixian method for example of any of claims 1-13.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710398014.7A CN108985804A (en) | 2017-05-31 | 2017-05-31 | Flow stage division and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710398014.7A CN108985804A (en) | 2017-05-31 | 2017-05-31 | Flow stage division and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108985804A true CN108985804A (en) | 2018-12-11 |
Family
ID=64501108
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710398014.7A Pending CN108985804A (en) | 2017-05-31 | 2017-05-31 | Flow stage division and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108985804A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111404835A (en) * | 2020-03-30 | 2020-07-10 | 北京海益同展信息科技有限公司 | Flow control method, device, equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103578010A (en) * | 2012-07-26 | 2014-02-12 | 阿里巴巴集团控股有限公司 | Method and device generating flow quality comparison parameters and advertisement billing method |
| WO2014189239A1 (en) * | 2013-05-20 | 2014-11-27 | 주식회사 애드오피 | Method and system of enhancing online contents value |
| CN105574741A (en) * | 2015-12-22 | 2016-05-11 | 北京奇虎科技有限公司 | Flow value assessment method, device, application method and system |
| CN105631711A (en) * | 2015-12-30 | 2016-06-01 | 合一网络技术(北京)有限公司 | Advertisement putting method and apparatus |
| WO2017004866A1 (en) * | 2015-07-09 | 2017-01-12 | 华为技术有限公司 | Method and device for determining advertisement value |
-
2017
- 2017-05-31 CN CN201710398014.7A patent/CN108985804A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103578010A (en) * | 2012-07-26 | 2014-02-12 | 阿里巴巴集团控股有限公司 | Method and device generating flow quality comparison parameters and advertisement billing method |
| WO2014189239A1 (en) * | 2013-05-20 | 2014-11-27 | 주식회사 애드오피 | Method and system of enhancing online contents value |
| WO2017004866A1 (en) * | 2015-07-09 | 2017-01-12 | 华为技术有限公司 | Method and device for determining advertisement value |
| CN105574741A (en) * | 2015-12-22 | 2016-05-11 | 北京奇虎科技有限公司 | Flow value assessment method, device, application method and system |
| CN105631711A (en) * | 2015-12-30 | 2016-06-01 | 合一网络技术(北京)有限公司 | Advertisement putting method and apparatus |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111404835A (en) * | 2020-03-30 | 2020-07-10 | 北京海益同展信息科技有限公司 | Flow control method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10192172B2 (en) | Methods and systems for predictive engine evaluation and replay of engine performance | |
| US10963541B2 (en) | Systems, methods, and apparatuses for implementing a related command with a predictive query interface | |
| US11176154B1 (en) | Collaborative dataset management system for machine learning data | |
| JP5662446B2 (en) | A learning system for using competitive evaluation models for real-time advertising bidding | |
| EP4062646A1 (en) | Systems and methods for identifying, tracking, and managing a plurality of social network users having predefined characteristcs | |
| CN111177111A (en) | Attribution modeling when executing queries based on user-specified segments | |
| US20130268520A1 (en) | Incremental Visualization for Structured Data in an Enterprise-level Data Store | |
| CN102722381B (en) | The technology of optimization and upgrading task | |
| CN105868243A (en) | Information processing method and apparatus | |
| CA2808096C (en) | Digital consumer data model and customer analytic record | |
| CN111797320B (en) | Data processing method, device, equipment and storage medium | |
| US9122995B2 (en) | Classification of stream-based data using machine learning | |
| CN103150696A (en) | Method and device for selecting potential customer of target value-added service | |
| CN104166713A (en) | Network service recommending method and device | |
| CN109345263A (en) | Predict the method and system of customer satisfaction | |
| CN113051317B (en) | Data mining model updating method, system, computer equipment and readable medium | |
| CN110189173B (en) | Advertisement generation method and device | |
| CN109190028A (en) | Activity recommendation method, apparatus, electronic equipment and storage medium based on big data | |
| US11698938B2 (en) | System and method for data event management system | |
| CN107203772B (en) | User type identification method and device | |
| CN110866698A (en) | Device for assessing service score of service provider | |
| US20230011954A1 (en) | Device, method, and system for business plan management | |
| Zhou et al. | Longitudinal impact of preference biases on recommender systems’ performance | |
| Proselkov et al. | Financial ripple effect in complex adaptive supply networks: an agent-based model | |
| CN108985804A (en) | Flow stage division and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20181221 Address after: Room 708, 7th floor, No. 10 Building, 30 Shixing Street, Shijingshan District, Beijing Applicant after: Beijing Xiaoxiong Bowang Technology Co., Ltd. Address before: 100085 Baidu Building, 10 Shangdi Tenth Street, Haidian District, Beijing Applicant before: BEIJING BAIDU NETCOM SCIENCE AND TECHNOLOGY Co.,Ltd. |
|
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181211 |