CN107944593A - A kind of resource allocation methods and device, electronic equipment - Google Patents
A kind of resource allocation methods and device, electronic equipment Download PDFInfo
- Publication number
- CN107944593A CN107944593A CN201710942224.8A CN201710942224A CN107944593A CN 107944593 A CN107944593 A CN 107944593A CN 201710942224 A CN201710942224 A CN 201710942224A CN 107944593 A CN107944593 A CN 107944593A
- Authority
- CN
- China
- Prior art keywords
- user
- resource
- order
- rate
- places
- 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
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
-
- 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
- G06Q30/0207—Discounts or incentives, e.g. coupons or rebates
- G06Q30/0224—Discounts or incentives, e.g. coupons or rebates based on user history
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0633—Lists, e.g. purchase orders, compilation or processing
- G06Q30/0635—Processing of requisition or of purchase orders
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Development Economics (AREA)
- Marketing (AREA)
- Accounting & Taxation (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Finance (AREA)
- Theoretical Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Physics & Mathematics (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Educational Administration (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
This application provides a kind of resource allocation methods, belong to field of computer technology, solve the problems, such as the existing distribution low of precision of resource allocation methods of the prior art the described method includes:Have that resource places an order rate and No Assets place an order rate by what the rate prediction model that places an order estimated user;Have that resource places an order rate and No Assets place an order rate based on described, determine the corresponding resource resources profit of the user;According to the order of the resource resources profit from high to low, resource allocation is carried out to relative users.Resource allocation methods disclosed in the embodiment of the present application, by carrying out resource allocation according to resource resources profit, can effectively improve the level of resources utilization, improve the precision of resource allocation.
Description
Technical field
This application involves field of computer technology, more particularly to a kind of resource allocation methods and device, electronic equipment.
Background technology
With the development of Internet technology, more and more network platforms are emerged in large numbers.In order to improve the user of platform or product
Viscosity, platform would generally distribute user the resource on some platforms, to strengthen registered users viscosity and attract new user.It is existing
Have in technology, generally according to historical data, be modeled based on the rate of placing an order, then determined by trained model to which user
Distribute resource.Alternatively, directly only meet that the user of preset threshold range distributes money to the rate of placing an order according to default resource allocation policy
Source, for example, only distributing resource to user of the rate of placing an order less than 0.5.However, the verification by long-term acquisition data, the prior art
In resource allocation methods at least there are resource allocation it is not accurate the problem of, i.e., resource, which is not allocated to, can really utilize the resource
User, be assigned to resource user do not use the resource.
As it can be seen that at least there is the defects of distribution precision is low in resource allocation methods of the prior art.
The content of the invention
The application provides a kind of resource allocation methods, and it is accurate to solve the existing distribution of resource allocation methods of the prior art
Spend the problem of low.
To solve the above-mentioned problems, in a first aspect, this application discloses a kind of resource allocation methods, including:
Have that resource places an order rate and No Assets place an order rate by what the rate prediction model that places an order estimated user;
Have that resource places an order rate and No Assets place an order rate based on described, determine the corresponding resource resources profit of the user;
According to the order of the resource resources profit from high to low, resource allocation is carried out to relative users.
Second aspect, this application discloses a kind of resource allocation device, including:
The rate of placing an order estimates module, has resource to place an order under rate and No Assets for estimate user by the rate prediction model that places an order
Single rate;
Resource resources profit determining module, for being placed an order rate and No Assets based on the resource that has that the rate of placing an order estimates that module estimates
Place an order rate, determines the corresponding resource resources profit of the user;
Resource distribution module, for the resource resources profit that is determined according to the resource resources profit determining module from high to low suitable
Relative users are carried out resource allocation by sequence.
The third aspect, the embodiment of the present application provide a kind of electronic equipment, including memory, processor and are stored in described
On memory and the computer program that can run on a processor, the processor realize this Shen when performing the computer program
Resource allocation methods that please be described in embodiment.
Fourth aspect, the embodiment of the present application provide a kind of computer-readable recording medium, are stored thereon with computer journey
The step of sequence, which realizes the resource allocation methods described in the embodiment of the present application when being executed by processor.
Resource allocation methods disclosed in the embodiment of the present application, have resource to place an order by what the rate prediction model that places an order estimated user
Rate and No Assets place an order rate;Then, have that resource places an order rate and No Assets place an order rate based on described, determine the corresponding money of the user
Source income;Finally, the order according to the resource resources profit from high to low, preferentially relative users high to resource resources profit carry out resource
Distribution, solves the problems, such as that the existing distribution precision of resource allocation methods of the prior art is low.In the embodiment of the present application
Disclosed resource allocation methods, by carrying out resource allocation according to resource resources profit, can effectively improve the level of resources utilization, be lifted
The precision of resource allocation.
Brief description of the drawings
, below will be in embodiment or description of the prior art in order to illustrate more clearly of the technical solution of the embodiment of the present application
Required attached drawing is briefly described, it should be apparent that, drawings in the following description are only some realities of the application
Example is applied, for those of ordinary skill in the art, without having to pay creative labor, can also be attached according to these
Figure obtains other attached drawings.
Fig. 1 is the resource allocation methods flow chart of the embodiment of the present application one;
Fig. 2 is the resource allocation methods flow chart of the embodiment of the present application two;
Fig. 3 is the resource allocation methods flow chart of the embodiment of the present application three;
Fig. 4 is the resource allocation methods flow chart of the embodiment of the present application four;
Fig. 5 is one of resource allocation device structure chart of the embodiment of the present application five;
Fig. 6 is the two of the resource allocation device structure chart of the embodiment of the present application five;
Fig. 7 is the three of the resource allocation device structure chart of the embodiment of the present application five.
Embodiment
Below in conjunction with the attached drawing in the embodiment of the present application, the technical solution in the embodiment of the present application is carried out clear, complete
Site preparation describes, it is clear that described embodiment is some embodiments of the present application, instead of all the embodiments.Based on this Shen
Please in embodiment, the every other implementation that those of ordinary skill in the art are obtained without creative efforts
Example, shall fall in the protection scope of this application.
Resource in the application can be virtual substance, such as electronic coupons, electronics coupons, integration, memory space;
Can also be actual contents, such as ticket, the product on gift or platform.The technical solution of the application is understood for the ease of reader,
In the specific embodiment of the application, illustrated using electronic coupons as resource, illustrate the concrete technical scheme of the application.
Embodiment one
A kind of resource allocation methods disclosed in the present application, as shown in Figure 1, this method includes:Step 100 is to step 120.
Step 100, have that resource places an order rate and No Assets place an order rate by what the rate prediction model that places an order estimated user.
The application first has to be gathered around according to the historical behavior data of user, user's representation data, user in the specific implementation
There is resource data to train the rate prediction model that places an order.It is when it is implemented, special according to the historical behavior data extracting section user of user
Sign (such as:The user behavior feature of default behavior), and, the lower forms data extraction label in user's history behavioral data
Data, and according to user's representation data extract part user characteristics (such as:User's Figure Characteristics), resource possessed according to user
Data extracting section user characteristics (such as:Resource characteristic);Then, the user characteristics of user and label data are formed into an instruction
Practice data, the training sample of the training data of multiple users as the rate prediction model that places an order, the trained rate prediction model that places an order.
It is special according to the user that user characteristics identical method extraction active user is extracted during training pattern during concrete application
Sign, wherein, the user characteristics of extraction includes:User behavior feature, user's Figure Characteristics and the resource characteristic of default behavior, so
Afterwards, the input using the user characteristics of extraction as the rate prediction model that places an order, estimate the active user have resource place an order rate and
No Assets place an order rate.When it is implemented, if the resource characteristic of extraction possesses the numerical value of resource for instruction active user,
Model output is has resource to place an order rate;If the resource characteristic of extraction does not have the numerical value of resource for instruction active user,
Model output places an order rate for No Assets.
Step 110, have that resource places an order rate and No Assets place an order rate based on described, determine that the corresponding resource of the user is received
Benefit.
When it is implemented, there is resource to place an order under rate and No Assets described in each user that can be estimated according to abovementioned steps
The difference of single rate, determines the resource resources profit of each user respectively.For example, by user u have the resource rate of placing an order subtract user u without money
Source places an order the difference of rate, the resource resources profit as user u.When it is implemented, other methods can also be used, according to user's
Have that resource places an order rate and the No Assets rate of placing an order determines the resource resources profit of a certain user, do not enumerated in the embodiment of the present application.
Step 120, relative users are carried out resource allocation by the order according to the resource resources profit from high to low.
After obtaining the resource resources profit of each user, user is arranged according to the order of the resource resources profit from high to low
Sequence;Then, sequentially preferentially the user high to resource resources profit distributes resource.For example, after obtaining the resource resources profit of each user, press
Sorted from front to back to user according to the order of the resource resources profit from high to low, then, according to the tandem point of user
With resource, until resource allocation is complete, or all users obtain resource.
Resource allocation methods disclosed in the embodiment of the present application, have resource to place an order by what the rate prediction model that places an order estimated user
Rate and No Assets place an order rate;Then, have that resource places an order rate and No Assets place an order rate based on described, determine the corresponding money of the user
Source income;Finally, relative users are carried out resource allocation, solved existing by the order according to the resource resources profit from high to low
The problem of existing distribution precision of resource allocation methods in technology is low.Resource allocation side disclosed in the embodiment of the present application
Method, by carrying out resource allocation according to resource resources profit, can effectively improve the level of resources utilization, improve the accurate of resource allocation
Degree.
Embodiment two
A kind of resource allocation methods disclosed in the present application, as shown in Fig. 2, this method includes:Step 200 is to step 230.
Step 200, the historical behavior data based on user, user's representation data, and user possess resource data, training
The rate that places an order prediction model.
When it is implemented, historical behavior data, user's representation data based on user, and user possess resource data,
The rate prediction model that places an order is trained, including:According to the first user characteristic data and the first label data, the training rate of placing an order estimates mould
Type;Wherein, first user characteristic data includes:The user behavior feature of default behavior, user's Figure Characteristics and for referring to
Show that user possesses the resource characteristic of resource situation;Whether first label data places an order for instruction user;The default row
For user behavior feature obtained according to the historical behavior data of the user, user's Figure Characteristics draw according to the user
As data acquisition, the resource characteristic possesses resource data according to the user and obtains.
The user behavior of the default behavior is characterized as the user characteristics from the historical behavior extracting data of user, its
In, the default behavior can include:One or more, the default behavior in behavior such as click on, browse, commenting on, placing an order
User behavior feature can include:User browses the number of a certain category, the number of a certain category product of comment, to a certain product
Lower single number of class product etc..
User's Figure Characteristics are the user characteristics extracted from user's representation data.User's representation data is retouched for user
State the essential information of user and the data of user behavior information.When it is implemented, user's representation data includes:The gender of user,
The personal information such as age, hobby, occupation.User's Figure Characteristics can include:In gender, age, hobby
It is one or more.When it is implemented, user's representation data can be obtained by the outer database of platform database or station, and further
User's Figure Characteristics are extracted from user's representation data.Drawn when it is implemented, being referred to method of the prior art from user
As extracting data user's Figure Characteristics, the present embodiment repeats no more.
The resource characteristic is the user characteristics for possessing resource data extraction according to user, and user possesses resource data can be with
For a part for user property, whether possess resource for instruction user, as whether user possesses electronic coupons.The resource
Whether feature possesses resource for instruction user, alternatively, which resource user possesses.
First label data is the label that the lower forms data in the historical behavior data according to user is extracted.For example,
A certain user has lower single act, then the first label data of the user is arranged to 1;A certain user does not descend single act, then the use
First label data at family is arranged to 0.
Then, the user behavior feature of the default behavior of user, user's Figure Characteristics and resource characteristic composition first is used
Family characteristic.Each user will obtain first user characteristic data.And by the first user characteristic data of the user and
First label data is as a training sample.Using two disaggregated models of supervised learning, such as logistic regression, random forest, god
Through network etc., modeling is fitted to the training sample data of all users, the probability that the label that model exports is 1, which is used as, to place an order
Rate.
Preferably, the user behavior feature of the default behavior, the historical behavior data based on user before predetermined time
Obtain;User's Figure Characteristics, user's representation data based on the predetermined time obtain;The resource characteristic, based on institute
The user for stating predetermined time possesses resource data acquisition;First label data, based on the user after the predetermined time
Lower forms data obtains.
When it is implemented, using some historical juncture t as boundary, such as t=2017 July 1, by the historical behavior number of user
According to two parts are divided into, a part is some period of another part after t moment within some period before t moment
It is interior.By taking t=2017 July 1 as an example, user's history behavioral data can be divided into A and B two parts:Part A includes 2017
The user's history behavioral data on January 1st, 31 days 1 June;Part B includes on July 7,2 days to 2017 July in 2017
User's history behavioral data.When it is implemented, the selection of the corresponding Period Length of A, B two parts user's history behavioral data needs
Determine as the case may be, the corresponding Period Length of part A data is more than the corresponding Period Length of part B data.For example, A portions
The corresponding Period Length of divided data is 6 months, and the corresponding Period Length of part B data is arranged to the term of validity of electronic coupons
(such as 7 days).
Then, user behavior spy of some statistics as default behavior is constructed according to part A user's history behavioral data
Sign.Such as user u browses the number of commodity under certain category, comments on business under certain category in statistics part A user's history behavioral data
The number of product and lower single number to commodity under certain category, the user behavior using these statistics as the default behavior of user u
Feature.Meanwhile user's Figure Characteristics are extracted from user's representation data of t moment, such as gender and age, as another part.
In addition, whether possess the resource characteristic that electronic coupons determine user further according to t moment user u.The resource characteristic is current
The fact that whether user possesses resource, such as:If the resource characteristic of extraction is 1, it indicates that active user possesses resource;
If the resource characteristic 0 of extraction, it indicates that active user does not have resource.Finally, by the user behavior feature of default behavior,
User's Figure Characteristics and resource characteristic, this three parts data constitute the first user characteristic data of user.Table 1 is exactly a letter
The first single user characteristic data.
Table 1, the first user characteristic data
Part B user's history behavioral data is recycled to construct first label data for each user.Described first
Whether label data places an order for instruction user within the part B data corresponding period.If when it is implemented, part B data
Middle lower single behavioral data there are user u, then can be arranged to 1 by the first label data of user u, otherwise by the of user u
One label data is arranged to 0, then can obtain the first label data as shown in Table 2.
User ID | Whether place an order |
123 | 0 |
124 | 1 |
125 | 1 |
126 | 0 |
Table 2, the first label data
When it is implemented, other methods can also be used, historical behavior data, user's representation data based on user, with
And user possesses resource data, the rate prediction model that places an order is trained.
For example, RNN, LSTM model of Series Modeling can be utilized to model the pervious behavior sequence of t moment.It is specifically walked
Suddenly it is that default behavior of the user in part A data is arranged in a sequence vector according to time order and function order, wherein, often
One vector be user on the day of behavioural characteristic list, vector in an element represent a kind of behavioural characteristic.Such as by user
Daily behavior vector is expressed as:[user is to the number of clicks of commodity in category 1, lower single of the user to commodity in category 1
Number, user is to the number of clicks of commodity in category 2, lower single number of the user to commodity in category 2].When it is implemented, behavior is special
Sign is not limited to 4 dimensions of the above.Assuming that the corresponding Period Length of part A data be 30 days, then can obtain as 30 to
Amount, forms sequence vector, and the sequence vector can be encoded to the vector of a regular length using models such as RNN, LSTM
h.It is then possible to the other users feature of vectorial h and user, such as user's Figure Characteristics, resource characteristic are connected into one completely
User characteristics, represented with characteristic vector sequence w, such as [h, user's Figure Characteristics 1 ..., user draw a portrait special user characteristics w=
Levy N, resource characteristic].Finally, user characteristics w is inputted to neutral net again, whether user is placed an order in part B data
The parameter and neural network parameter of output of the behavioral data as model, training and optimization RNN/LSTM.In this model, RNN/
The parameter and neural network parameter of LSTM is using combined optimization end to end.
Step 210, have that resource places an order rate and No Assets place an order rate by what the rate prediction model that places an order estimated user.
When needing to distribute resource, user's row of the default behavior of user is determined according to the historical behavior data of user first
It is characterized, user's Figure Characteristics and resource characteristic.When it is implemented, the pervious user's history behavioral data of current point in time is made
To extract the source data of first user characteristic data, according to training pattern when extracts the identical side of the first user characteristic data
User behavior feature, user's Figure Characteristics and the resource characteristic of the default behavior of method extraction active user.Then, by the institute of extraction
The input of user behavior feature, user's Figure Characteristics and the resource characteristic of default behavior as the rate prediction model that places an order is stated, is estimated
The active user's has that resource places an order rate or No Assets place an order rate.If when it is implemented, the resource characteristic is arranged to
Indicate that active user possesses the numerical value of resource, then model output is has resource to place an order rate;If the resource characteristic is arranged to
Instruction active user does not have the numerical value of resource, then model output places an order rate for No Assets.
Step 220, have that resource places an order rate and No Assets place an order rate based on described, determine that the corresponding resource of the user is received
Benefit.
When it is implemented, having that resource places an order rate and No Assets place an order rate based on described, the corresponding resource of the user is determined
Income, including:Have that resource places an order rate and No Assets place an order the difference of rate according to user, determine that the resource of the user is received
Benefit.Such as the resource resources profit of a certain user is determined by equation 1 below:Profit (u)=OrderWithResource (u)-
OrderWithoutResource(u);(formula 1)
In formula 1, Profit (u) represents the resource resources profit of user u, and OrderWithResource (u) represents user u's
There is resource to place an order rate, OrderWithoutResource (u) represents that the No Assets of user u place an order rate.Wherein,
OrderWithResource (u) and OrderWithoutResource (u) estimate to obtain by the rate prediction model that places an order.Specifically
During implementation, other methods can also be used, have that resource places an order rate and the No Assets rate of placing an order determines a certain user according to user
Resource resources profit, for example, Profit (u)=OrderWithResource (u)/OrderWithoutResource (u), this Shen
It please not enumerated in embodiment.
To there is the historical behavior data instance of 5 users in user's history behavioral data, 5 users are expressed as:
U1, u2, u3, u4 and u5, after the first user characteristic data based on historical behavior data extraction user u1 to u5, pass through
What the rate prediction model of placing an order estimated user u1 to u5 respectively has that resource places an order rate and No Assets place an order rate, and is calculated according to formula 1
The resource resources profit of user u1 to u5, obtained resource resources profit are expressed as:Profit(u1)、Profit(u2)、Profit
(u3), Profit (u4) and Profit (u5).
Step 230, relative users are carried out resource allocation by the order according to the resource resources profit from high to low.
The resource resources profit for each user for possessing historical behavior data can be obtained by formula 1.Then, according to the money
The order of source income from high to low is ranked up user, and sequentially preferentially the user high to resource resources profit distributes resource.With with
There is the historical behavior data instance of 5 users in the historical behavior data of family, the resource resources profit of user u1 to u5 is calculated according to formula 1
It is expressed as:Profit (u1), Profit (u2), Profit (u3), Profit (u4) and Profit (u5).Then, according to
The orders of Profit () from high to low sequentially carry out resource allocation to relative users.Assuming that resource resources profit sequence is:Profit
(u1)>Profit(u2)>Profit(u3)>Profit(u5)>Profit (u4), then first to the corresponding users of Profit (u1)
U1 carries out resource allocation, then, then carries out resource allocation to the corresponding user u2 of Profit (u2), finally right to Profit (u4)
The user u4 answered carries out resource allocation.If inadequate resource, to distribute to all users, the low user of resource resources profit will distribute not
To resource.
Resource allocation methods disclosed in the embodiment of the present application, pass through the historical behavior data based on user, user's portrait number
According to and user possesses resource data, and training places an order rate prediction model, and has money by what the rate prediction model that places an order estimated user
Rate that source places an order rate and No Assets place an order;Then, have that resource places an order rate and No Assets place an order rate based on described, determine the user couple
The resource resources profit answered;Finally, relative users are carried out resource allocation, solved by the order according to the resource resources profit from high to low
The problem of existing distribution precision of resource allocation methods of the prior art is low.Resource disclosed in the embodiment of the present application
Distribution method, user behavior feature, user's Figure Characteristics and resource characteristic based on user train place an order rate prediction model, and base
In having of estimating of the rate prediction model of placing an order, resource places an order rate and the No Assets rate of placing an order determines resource resources profit, last according to resource resources profit
Resource allocation is carried out, when carrying out resource allocation, has taken into full account historical behavior, the user's portrait of user, and user is to money
The service condition in source, effectively improves the level of resources utilization, improves the precision of resource allocation.
Embodiment three
A kind of resource allocation methods disclosed in the present application, as shown in figure 3, this method includes:Step 300 is to step 350.
Step 300, the historical behavior data based on user, user's representation data, and user possess resource data, training
The rate that places an order prediction model.
Historical behavior data, user's representation data based on user, and user possess resource data, train the rate that places an order pre-
Model is estimated referring to embodiment two, and details are not described herein again.
Step 310, historical behavior data and user's representation data based on the user for possessing resource, the lower single price of training are pre-
Estimate model.
When it is implemented, historical behavior data and user's representation data based on the user for possessing resource, the lower unit price of training
Lattice prediction model, including:According to the second user characteristic and the second label data of the user for possessing resource, the lower unit price of training
Lattice prediction model;Wherein, the second user characteristic includes:The corresponding price feature of default behavior, user's Figure Characteristics;
Second label data is used for lower single price of instruction user;The corresponding user characteristics of the default behavior is according to the user
Historical behavior data obtain, user's Figure Characteristics obtain according to user's representation data.
The user behavior of the default behavior is characterized as the user characteristics from the historical behavior extracting data of user, its
In, the default behavior can include:One or more, the default behavior in behavior such as click on, browse, commenting on, placing an order
Corresponding price feature can include:The price for a certain category that user browses, the price of a certain category product of comment, user
Price of a certain category product of purchase etc..
User's Figure Characteristics are the user characteristics extracted from user's representation data.User's representation data is retouched for user
State the essential information of user and the data of user behavior information.When it is implemented, user's representation data includes:The gender of user,
The personal information such as age, hobby, occupation.User's Figure Characteristics can include:In gender, age, hobby
It is one or more.When it is implemented, user's representation data can be obtained by the outer database of platform database or station, and further
User's Figure Characteristics are extracted from user's representation data.Drawn when it is implemented, being referred to method of the prior art from user
As extracting data user's Figure Characteristics, the present embodiment repeats no more.
Second label data is the label that the lower forms data in the historical behavior data according to user is extracted.For example,
A certain user has lower single act, then the user place an order the product of purchase price be 25, then the first label data of the user set
It is set to 25.If a certain user does not descend single act, not using the data of the user as training during single price estimation model under training
Data.
Then, the corresponding price feature of default behavior of user, user's Figure Characteristics are formed into second user characteristic.
Each user will obtain a second user characteristic.And by the second user characteristic and the second label data of the user
As a training sample.Using two disaggregated models of supervised learning, such as logistic regression, random forest, neutral net, to institute
The training sample data for having user are fitted modeling, and model output is lower single price of a certain user.
Preferably, the corresponding price feature of the default behavior, the historical behavior data based on user before predetermined time
Obtain;User's Figure Characteristics, user's representation data based on the predetermined time obtain;Second label data, base
Forms data obtains under the user after the predetermined time.
When it is implemented, using some historical juncture t as boundary, such as t=2017 July 1, by the historical behavior number of user
According to two parts are divided into, a part is some period of another part after t moment within some period before t moment
It is interior.By taking t=2017 July 1 as an example, user's history behavioral data can be divided into A and B two parts:Part A includes 2017
The user's history behavioral data on January 1st, 31 days 1 June;Part B includes on July 7,2 days to 2017 July in 2017
User's history behavioral data.When it is implemented, the selection of the corresponding Period Length of A, B two parts user's history behavioral data needs
Determine as the case may be, the corresponding Period Length of part A data is more than the corresponding Period Length of part B data.For example, A portions
The corresponding Period Length of divided data is 6 months, and the corresponding Period Length of part B data is arranged to the term of validity of electronic coupons
(such as 7 days).
Then, price feature of some statistics as default behavior is constructed according to part A user's history behavioral data.Example
Such as count user u in part A user's history behavioral data and browse the average price of commodity under certain category, business under certain category of comment
The average price of the average price of product and certain category commodity of purchase, the valency using these statistics as the default behavior of user u
Lattice feature.Meanwhile user's Figure Characteristics are extracted from user's representation data of t moment, such as gender and age, as another portion
Point.Finally, by the price feature of default behavior, user's Figure Characteristics, the second user that this two partial data constitutes user is special
Levy data.Table 3 is exactly a simple second user characteristic.
Table 3, second user characteristic
Part B user's history behavioral data is recycled to construct second label data for each user.Described second
Label data is used for lower single price of the instruction user within the part B data corresponding period.If when it is implemented, part B number
According to middle lower single behavioral data there are user u, then the second label data of user u can be arranged to being averaged for the user and placed an order
Price, otherwise abandons the second user characteristic of user u, and the data of the user will be not used in as the lower single price estimation of training
The training sample of model.The second label data as shown in table 4 can then be obtained.
User ID | Lower list average price |
123 | 100 |
124 | 56 |
125 | 200 |
Table 4, the second label data
When it is implemented, other methods can also be used, and historical behavior data, user's representation data based on user, instruction
Practice the rate prediction model that places an order.For example, RNN, LSTM model of Series Modeling can be utilized to build the pervious behavior sequence of t moment
Mould.The pervious behavior sequence of t moment is modeled using RNN, LSTM model of Series Modeling, the lower single price estimation model of training
Referring to the specific method that the rate prediction model that places an order is trained in embodiment two, difference is specific method:Utilize Series Modeling
RNN, LSTM model model the pervious behavior sequence of t moment, and under training during single price estimation model, the element of sequence vector is
Price feature and user's Figure Characteristics, the output terminal of model are the prices in the second data label.
Step 320, have that resource places an order rate and No Assets place an order rate by what the rate prediction model that places an order estimated user.
Have that resource places an order rate and No Assets place an order the embodiment of rate by what the rate prediction model that places an order estimated user
Referring to embodiment two, details are not described herein again.
Step 330, lower single price of lower single price estimation model pre-estimating user is passed through.
When needing to distribute resource, the second user characteristic of user is determined according to the historical behavior data of user first
According to the second user characteristic includes:The corresponding price feature of default behavior, user's Figure Characteristics.When it is implemented, handle
The pervious user's history behavioral data of current point in time is as the source data for extracting the second user characteristic, according to training
The default behavior of the identical method extraction active user of the corresponding price feature of default behavior, user's Figure Characteristics is extracted during model
Corresponding price feature, user's Figure Characteristics.Then, by the corresponding price feature of default behavior, the user's Figure Characteristics of extraction
As the input of lower single price estimation model, lower single price of the active user is estimated.
Step 340, have that resource places an order rate and No Assets place an order rate based on described, determine that the corresponding resource of the user is received
Benefit.
When it is implemented, having that resource places an order rate and No Assets place an order rate based on described, the corresponding resource of the user is determined
Income, including:Have that resource places an order rate and No Assets place an order rate, and lower single price of the user according to user, really
The resource resources profit of the fixed user.Preferably, have that resource places an order rate and No Assets place an order rate according to user, and it is described
Lower single price of user, determines the resource resources profit of the user, including:There is resource to place an order rate and No Assets according to user
The product of the difference for the rate that places an order and lower single price of the user, determines the resource resources profit of the user.Such as pass through following public affairs
Formula 2 determines the resource resources profit of a certain user:
Profit (u)=(OrderWithResource (u)-OrderWithoutResource (u)) * price (u);
(formula 2)
In formula 2, Profit (u) represents the resource resources profit of user u, and OrderWithResource (u) represents user u's
There is resource to place an order rate, OrderWithoutResource (u) represents that the No Assets of user u place an order rate, and price (u) represents user u
Estimating under single price.Wherein, OrderWithResource (u) and OrderWithoutResource (u) are pre- by the rate of placing an order
Estimate model pre-estimating to obtain;Price (u) is obtained by lower single price estimation model pre-estimating.When it is implemented, other can also be used
Method, resource places an order rate to having based on user and the No Assets rate of placing an order determines the resource resources profit of a certain user, for example, Profit
(u)=OrderWithResource (u)/OrderWithoutResource (u) * price (u), differs in the embodiment of the present application
One enumerates.
Still to there is the historical behavior data instance of 5 users in user's history behavioral data, 5 users are represented respectively
For:U1, u2, u3, u4 and u5, are expressed as by the resource resources profit that user u1 to u5 is calculated according to formula 2:Profit
(u1), Profit (u2), Profit (u3), Profit (u4) and Profit (u5).
Step 350, relative users are carried out resource allocation by the order according to the resource resources profit from high to low.
The resource resources profit for each user for possessing historical behavior data can be obtained by formula 2.Then, according to the money
The order of source income from high to low is ranked up user, and sequentially preferentially the user high to resource resources profit distributes resource.According to
The order of the resource resources profit from high to low, carries out relative users the embodiment of resource allocation referring to embodiment two,
Details are not described herein again.According to the order of the resource resources profit from high to low, relative users are carried out with resource allocation, resource resources profit is high
User will be assigned to resource a little, if inadequate resource, to distribute to all users, the low user of resource resources profit will distribute not
To resource.
When it is implemented, placing an order rate prediction model and lower single price estimation model is typically training in advance, this Shen under line
Please the sequencing for training place an order rate prediction model and lower single price estimation model is not limited.When carrying out resource allocation,
The application has that resource places an order rate and No Assets place an order and rate and estimate the sequencing of lower single price and do not limit to estimate user.
Resource allocation methods disclosed in the embodiment of the present application, pass through the historical behavior data based on user, user's portrait number
According to and user possesses resource data, and training places an order rate prediction model, and has money by what the rate prediction model that places an order estimated user
Rate that source places an order rate and No Assets place an order;And historical behavior data based on user, user's representation data, the lower single price of training
Prediction model, and the lower single price for passing through lower single price estimation model pre-estimating user;Then, based on it is described have resource place an order rate and
No Assets place an order rate, and lower single price, determine the corresponding resource resources profit of the user;Finally, according to the resource resources profit by
Relative users are carried out resource allocation by high to Low order, solve the existing distribution of resource allocation methods of the prior art
The problem of precision is low.Resource allocation methods disclosed in the embodiment of the present application, are placed an order by regarding lower single price as training
The source data of price estimation model, can improve the accuracy rate of lower single price estimation, meanwhile, provided by combining lower single price estimation
Source income, can further lift the reliability for the resource resources profit estimated so that when carrying out resource allocation according to resource resources profit, fill
The price feature for considering the historical behavior of user, user's portrait, and user is divided to effectively improve money to the service condition of resource
Source utilization ratio, improves the precision of resource allocation.
Example IV
A kind of resource allocation methods disclosed in the present application, as shown in figure 4, this method includes:Step 400 is to step 460.
Step 400, the historical behavior data based on user, user's representation data, and user possess resource data, training
The rate that places an order prediction model.
When it is implemented, platform may only have a kind of resource, it is also possible to there are multiple resources, for example, it may be possible to which there are electronics
Coupons and completely subtracting reward voucher, historical behavior data, user's representation data based on user, and user possess resource data,
Train then needs to consider the resource characteristic of every kind of resource when placing an order rate prediction model.When there is multiple resources, resource characteristic is specific
Whether possess resource for instruction user, and which kind of resource possessed.By taking platform has 2 kinds of resources as an example, resource characteristic can wrap
Include three dimensions:Whether first dimension is used for instruction user without any resource;Whether second dimension is used for instruction user
Possess resource 1;Whether the 3rd dimension possesses resource 2 for instruction user.The extracting mode of resource characteristic referring to embodiment two,
Details are not described herein again.When it is implemented, if first dimension is 1 in the resource characteristic of extraction, it indicates that active user
There is no resource;If first dimension is 0 in the resource characteristic of extraction, second dimension is 1, it indicates that active user gathers around
There is resource and possess resource 1.
Historical behavior data extraction based on user presets the embodiment of the user behavior feature of behavior, based on use
The embodiment of family representation data extraction user's Figure Characteristics is referring to embodiment two, and details are not described herein again.
By taking platform has 2 kinds of resources as an example, historical behavior data, user's representation data based on user, and user possess
The form for the first user characteristic data that resource data obtains can be as shown in table 5.
Table 5, the first user characteristic data
Historical behavior data based on user extract the embodiment of the first label data referring to embodiment two, herein
Repeat no more.
Then, the user behavior feature of the default behavior of user, user's Figure Characteristics and resource characteristic composition first is used
Family characteristic.Each user will obtain first user characteristic data.And by the first user characteristic data of the user and
First label data is as a training sample.Using two disaggregated models of supervised learning, such as logistic regression, random forest, god
Through network etc., modeling is fitted to the training sample data of all users, the probability that the label that model exports is 1, which is used as, to place an order
Rate.
Same way, it is also possible to using other methods, historical behavior data, user's representation data based on user, and user
Possess resource data, train the rate prediction model that places an order.For example, can utilize Series Modeling RNN, LSTM model to t moment with
Preceding behavior sequence modeling.The pervious behavior sequence of t moment is modeled using RNN, LSTM model of Series Modeling, training places an order
The specific method of rate prediction model participates in embodiment two, and details are not described herein again.
Step 410, historical behavior data and user's representation data based on the user for possessing resource, the lower single price of training are pre-
Estimate model.
Historical behavior data and user's representation data based on the user for possessing resource, the lower single price estimation model of training
Embodiment is referring to embodiment three, and details are not described herein again.
Step 420, have that resource places an order rate and No Assets place an order rate by what the rate prediction model that places an order estimated user.
By the rate prediction model that places an order estimate user have resource place an order rate and No Assets place an order rate when, it is necessary to input treat it is pre-
Estimate user behavior feature, user's Figure Characteristics and the resource characteristic of the default behavior of user.
When platform is there are during multiple resources, the user behavior feature of the default behavior of user, user's Figure Characteristics it is specific
Extracting mode is identical with embodiment two, and details are not described herein again.
When platform is there are during multiple resources, whether the resource characteristic of user possesses resource for instruction user, and possesses
Which kind of resource.Still by taking platform has 2 kinds of resources as an example, the resource characteristic at the current time of extraction can be as shown in table 6.
Table 6, including resource characteristic during multiple resources
Wherein, whether a dimension possesses resource for instruction user;Whether one dimension possesses money for instruction user
Source 1;Whether one dimension possesses resource 2 for instruction user.
Then, by the user behavior feature of the default behavior of extraction, user's Figure Characteristics, and resource characteristic is set to make
For the input for the rate prediction model that places an order, that estimates the active user has that resource places an order rate or No Assets place an order rate.Specific implementation
When, if setting the resource characteristic to possess resource for instruction active user and possess the numerical value of resource 1, model output is to have
The rate that places an order of resource 1;If setting the resource characteristic to possess resource for instruction active user and possess the numerical value of resource 2,
Model output is the rate that places an order for having resource 2;If the resource characteristic is set not have the number of any resource for instruction active user
Value, then model output place an order rate for No Assets.
Have that resource places an order rate and No Assets place an order after rate by what the rate prediction model that places an order estimated user, user will be obtained
The resource that has for possessing every kind of resource places an order rate, and the No Assets of user place an order rate.
When it is implemented, when there are 2 kinds of resources in platform, resource characteristic can also be represented using two dimensions, such as:
Certificate 1 is put 1, certificate 2 is set to 0, and the rate that places an order that user only possesses resource 1 is estimated in expression;Certificate 1 is set to 0, certificate 2 puts 1, and user is estimated in expression
Only possess the rate that places an order of resource 2;Certificate 1 is set to 0, certificate 2 is set to 0, and user is estimated in expression does not have the rate that places an order of any resource.Likewise,
When training places an order rate prediction model, the resource characteristic for also extracting two dimensions is used to train the rate prediction model that places an order.
Step 430, lower single price of lower single price estimation model pre-estimating user is passed through.
By the embodiment of lower single price of lower single price estimation model pre-estimating user referring to embodiment three, herein
Repeat no more.
When platform is there are during multiple resources, when it is implemented, having that resource places an order rate and No Assets place an order rate based on described, really
Determine the corresponding resource resources profit of the user, including:There is resource to place an order rate according to user and No Assets place an order rate, Yi Jisuo
Lower single price of user and the cost of each resource are stated, determines the single resource resources profit of every kind of resource of the user;By maximum
The single resource resources profit is as the corresponding resource resources profit of the user.In the present embodiment, it is being determined that user's have resource to place an order
Rate and No Assets place an order rate, and after lower single price of the user, it is first determined every kind of resource of the user it is single
Resource resources profit.
Step 440, have that resource places an order rate and No Assets place an order rate, and the lower unit price of the user according to user
The cost of lattice and each resource, determines the single resource resources profit of the every kind of resource of the user.
Preferably, have that resource places an order rate and No Assets place an order rate, and the lower unit price of the user according to user
The cost of lattice and each resource, determining the single resource resources profit of every kind of resource of the user includes:Determine that the described of user has money
The difference for rate that source places an order rate and No Assets place an order;According to the product and single resource of the difference and lower single price of the user
Cost ratio, determine the single resource resources profit of the user.Such as each resource of a certain user is determined by equation 3 below
Single resource resources profit:
Profit ' (u, v)=(OrderWithResource (u, v)-OrderWithoutResource (u)) * price
(u)/cos(v);(formula 3)
In formula 3, Profit ' (u, v) represents the single resource resources profit of the resource v of user u, OrderWithResource
(u, v) represents the rate that places an order having during resource v of user u, and OrderWithoutResource (u) is represented under the No Assets of user u
Single rate, price (u) represent single price under the estimating of user u, and cos (v) is the cost of resource v.Wherein,
OrderWithResource (u, v) and OrderWithoutResource (u) estimates to obtain by the rate prediction model that places an order;
Price (u) is obtained by lower single price estimation model pre-estimating;Cos (v) is the constant that platform is set, such as electronic offers
Certificate, the value of cos (v) include face amount and short message cost of reward voucher etc..When it is implemented, other methods can also be used,
Resource places an order rate to having based on user and the No Assets rate of placing an order determines the resource resources profit of a certain user, for example,
Profit ' (u, v)=(OrderWithResource (u, v)/OrderWithoutResource (u)) * price
(u)/cos (v), does not enumerate in the embodiment of the present application.
Step 450, using the maximum single resource resources profit as the corresponding resource resources profit of the user.
If platform has two spike-type cultivars, such as resource 1 and resource 2, then can calculate respectively when user u possesses resource 1
Single resource resources profit Profit1' and single resource resources profit Profits of user u when possessing resource 22’.Then, by described in maximum
Single resource resources profit is used as this as the corresponding resource resources profit of the user, using the corresponding resource of the maximum single resource resources profit
User's resource the most matched.For example, work as Profit1’>Profit2' when, by Profit1' resource resources profit as user u, money
Most matching resource of the source 1 as user u.
Using same method, the money of each user on platform can be determined respectively by repeating step 440 and step 450
Source income.
Step 460, relative users are carried out resource allocation by the order according to the resource resources profit from high to low.
After the resource resources profit that each user is determined, user is carried out according to the order of the resource resources profit from high to low
Sequence, and sequentially preferentially the user high to resource resources profit distributes resource.According to the order of the resource resources profit from high to low, to phase
Using the embodiment of family progress resource allocation referring to embodiment two, details are not described herein again.According to the resource resources profit by
Relative users are carried out resource allocation, the high user of resource resources profit will be assigned to resource a little, if resource by high to Low order
It is not enough to distribute to all users, the low user of resource resources profit will be distributed less than resource.The resource for distributing to user is the user
Most matching resource.
When it is implemented, placing an order rate prediction model and lower single price estimation model is typically training in advance, this Shen under line
Please the sequencing for training place an order rate prediction model and lower single price estimation model is not limited.When carrying out resource allocation,
The application has that resource places an order rate and No Assets place an order and rate and estimate the sequencing of lower single price and do not limit to estimate user.
Resource allocation methods disclosed in the embodiment of the present application, the situation by possessing various resources based on user are extracted different
The resource characteristic of dimension, and user behavior feature and user's Figure Characteristics are combined, and one-state trains the rate that places an order pre- under user
Estimate model, resource places an order rate and user does not have No Assets during resource to place an order for estimating having when user possesses different resource
Rate;Then, have that resource places an order rate and No Assets place an order rate, and lower single price, resources costs with reference to described, determine single resource
Income, and using maximum single resource resources profit as the corresponding resource resources profit of the user;Finally, according to the resource resources profit by height to
Relative users are carried out resource allocation by low order, and it is accurate to solve the existing distribution of resource allocation methods of the prior art
Spend the problem of low.Resource allocation methods disclosed in the embodiment of the present application, are used as by the situation that user is possessed to each resource
A part for resource characteristic, as the foundation of the single resource resources profit of definite user, and using maximum single resource resources profit as the use
The resource resources profit at family, takes into full account the situation of Profit of different resource, can further lift the reliability for the resource resources profit estimated,
The level of resources utilization is effectively improved, improves the precision of resource allocation.
Embodiment five
A kind of resource allocation device disclosed in the present application, as shown in figure 5, described device includes:
The rate of placing an order estimates module 510, and the resource that has for estimating user by the rate prediction model that places an order places an order rate and without money
Source places an order rate;
Resource resources profit determining module 520, for based on the rate of placing an order estimate that module 510 estimates have resource place an order rate and
No Assets place an order rate, determine the corresponding resource resources profit of the user;
Resource distribution module 530, for according to the resource resources profit determining module 520 determine resource resources profit from high to low
Order, to relative users carry out resource allocation.
Resource allocation device disclosed in the embodiment of the present application, has resource to place an order by what the rate prediction model that places an order estimated user
Rate and No Assets place an order rate;Then, have that resource places an order rate and No Assets place an order rate based on described, determine the corresponding money of the user
Source income;Finally, relative users are carried out resource allocation, solved existing by the order according to the resource resources profit from high to low
The problem of existing distribution precision of resource allocation methods in technology is low.Resource allocation side disclosed in the embodiment of the present application
Method, by carrying out resource allocation according to resource resources profit, can effectively improve the level of resources utilization, improve the accurate of resource allocation
Degree.
Optionally, as shown in fig. 6, the resource resources profit determining module 520 includes:
First resource income determination unit 5201, for thering is resource to place an order rate according to user and No Assets place an order rate
Difference, determine the resource resources profit of the user.
Have that resource places an order rate and No Assets place an order the difference of rate according to user, determine the resource resources profit of the user
Specific embodiment referring to embodiment two, details are not described herein again.
Optionally, as shown in fig. 6, described device further includes:
The rate that places an order prediction model training module 540, for the historical behavior data based on user, user's representation data, with
And user possesses resource data, the rate prediction model that places an order is trained.
Optionally, the rate prediction model training module 540 that places an order is further used for:
According to the first user characteristic data and the first label data, the rate prediction model that places an order is trained;
Wherein, first user characteristic data includes:User behavior feature, user's Figure Characteristics and the use of default behavior
Possess the resource characteristic of resource situation in instruction user;Whether first label data places an order for instruction user;It is described pre-
If the user behavior feature of behavior is obtained according to the historical behavior data of the user, user's Figure Characteristics are according to the use
Family representation data obtains, and the resource characteristic possesses resource data according to the user and obtains.
Preferably, the user behavior feature of the default behavior, the historical behavior data based on user before predetermined time
Obtain;User's Figure Characteristics, user's representation data based on the predetermined time obtain;The resource characteristic, based on institute
The user for stating predetermined time possesses resource data acquisition;First label data, based on the user after the predetermined time
Lower forms data obtains.
Resource allocation device disclosed in the embodiment of the present application, user behavior feature, user's Figure Characteristics based on user
The rate prediction model that places an order is trained with resource characteristic, and has resource to place an order under rate and No Assets based on what the rate prediction model of placing an order was estimated
Single rate determines resource resources profit, last to carry out resource allocation according to resource resources profit, when carrying out resource allocation, has taken into full account user
Historical behavior, user's portrait, and user effectively improves the level of resources utilization, improves resource to the service condition of resource
The precision of distribution.
Optionally, as shown in fig. 7, described device further includes:
Lower list price estimation module 550, for lower single price by lower single price estimation model pre-estimating user;
The resource resources profit determining module 520 further includes:
Secondary resource income determination unit 5202, for thering is resource to place an order rate according to user and No Assets place an order
Rate, and lower single price of the user, determine the resource resources profit of the user.Have according to user resource place an order rate and
No Assets place an order rate, and lower single price of the user, determine the user resource resources profit specific embodiment referring to
Embodiment three, details are not described herein again.
The Secondary resource income determination unit 5202 is further used for:
Have that resource places an order rate and No Assets place an order the difference of rate and lower single price of the user according to user
Product, determines the resource resources profit of the user.
Optionally, as shown in fig. 7, described device further includes:
Lower list price estimation model training module 560, for historical behavior data and use based on the user for possessing resource
Family representation data, the lower single price estimation model of training.
Optionally, lower single price estimation model training module 560 is further used for:
According to the second user characteristic and the second label data of the user for possessing resource, the lower single price estimation mould of training
Type;
Wherein, the second user characteristic includes:The corresponding price feature of default behavior, user's Figure Characteristics;Institute
State lower single price that the second label data is used for instruction user;The corresponding user characteristics of the default behavior is according to the user's
Historical behavior data obtain, and user's Figure Characteristics are obtained according to user's representation data.
Preferably, the corresponding price feature of the default behavior, the historical behavior data based on user before predetermined time
Obtain;User's Figure Characteristics, user's representation data based on the predetermined time obtain;Second label data, base
Forms data obtains under the user after the predetermined time.
When platform possesses multiple resources, by lower single price estimation module 550 estimate user lower single price it
Afterwards, optionally, as shown in fig. 7, the resource resources profit determining module 520 further includes:
Information resources income determination unit 5203, for thering is resource to place an order rate according to user and No Assets place an order
Rate, and lower single price of the user and the cost of each resource, determine the single resource resources profit of every kind of resource of the user;
The information resources income determination unit 5203, is additionally operable to using the maximum single resource resources profit as the user
Corresponding resource resources profit.
Optionally, the information resources income determination unit 5203 is further used for:
Determine that the described of user has that resource places an order rate and No Assets place an order the difference of rate;According to the difference and the user
Lower single price product and the cost of single resource ratio, determine the single resource resources profit of the user.
The specific embodiment of information resources income determination unit is referring to example IV, and details are not described herein again
Resource allocation device disclosed in the embodiment of the present application, by regarding lower single price as the lower single price estimation mould of training
The source data of type, can improve the accuracy rate of lower single price estimation, meanwhile, can by combining lower single price estimation resource resources profit
With the further reliability for lifting the resource resources profit estimated so that when carrying out resource allocation according to resource resources profit, take into full account
Price feature, the user's portrait of the historical behavior of user, and user effectively improve utilization of resources effect to the service condition of resource
Rate, improves the precision of resource allocation.
Further, a part for resource characteristic is used as by the way that user to be possessed to the situation of each resource, as definite user
The foundation of single resource resources profit, and using maximum single resource resources profit as the resource resources profit of the user, take into full account different resource
Situation of Profit, can further lift the reliability for the resource resources profit estimated, effectively improve the level of resources utilization, improve money
The precision of source distribution.
Correspondingly, disclosed herein as well is a kind of electronic equipment, including memory, processor and it is stored in the memory
Computer program that is upper and can running on a processor, the processor are realized when performing the computer program as the application is real
Example one is applied to the resource allocation methods described in example IV.The electronic equipment can be mobile terminal, smart phone, navigation
Instrument, personal digital assistant, tablet computer etc..
Disclosed herein as well is a kind of computer-readable recording medium, computer program is stored thereon with, which is located
Manage the step of realizing the resource allocation methods as described in the embodiment of the present application one to example IV when device performs.
Each embodiment in this specification is described by the way of progressive, what each embodiment stressed be with
The difference of other embodiment, between each embodiment identical similar part mutually referring to.For device embodiment
For, since it is substantially similar to embodiment of the method, so description is fairly simple, referring to the portion of embodiment of the method in place of correlation
Defend oneself bright.
A kind of resource allocation methods and device provided above the application are described in detail, tool used herein
Body example is set forth the principle and embodiment of the application, and the explanation of above example is only intended to help to understand this Shen
Method and its core concept please;Meanwhile for those of ordinary skill in the art, according to the thought of the application, specific real
There will be changes in mode and application range are applied, in conclusion this specification content should not be construed as the limit to the application
System.
Through the above description of the embodiments, those skilled in the art can be understood that each embodiment can
Realized by the mode of software plus required general hardware platform, naturally it is also possible to pass through hardware realization.Based on such reason
Solution, the part that above-mentioned technical proposal substantially in other words contributes the prior art can be embodied in the form of software product
Come, which can store in a computer-readable storage medium, such as ROM/RAM, magnetic disc, CD, including
Some instructions are used so that a computer equipment (can be personal computer, server, or network equipment etc.) performs respectively
Method described in some parts of a embodiment or embodiment.
Claims (11)
- A kind of 1. resource allocation methods, it is characterised in that including:Have that resource places an order rate and No Assets place an order rate by what the rate prediction model that places an order estimated user;Have that resource places an order rate and No Assets place an order rate based on described, determine the corresponding resource resources profit of the user;According to the order of the resource resources profit from high to low, resource allocation is carried out to relative users.
- 2. according to the method described in claim 1, it is characterized in that, described have that resource places an order rate and No Assets place an order based on described Rate, the step of determining the user corresponding resource resources profit, including:Have that resource places an order rate and No Assets place an order the difference of rate according to user, determine the resource resources profit of the user.
- 3. according to the method described in claim 1, it is characterized in that, described have that resource places an order rate and No Assets place an order based on described Rate, before the step of determining the user corresponding resource resources profit, further includes:Pass through lower single price of lower single price estimation model pre-estimating user;It is described to have that resource places an order rate and No Assets place an order rate based on described, the step of determining the user corresponding resource resources profit, Including:Have that resource places an order rate and No Assets place an order rate, and lower single price of the user according to user, determine described The resource resources profit of user.
- 4. according to the method described in claim 3, it is characterized in that, described have resource to place an order rate and without money according to user Source places an order rate, and lower single price of the user, the step of determining the resource resources profit of the user, including:Have that resource places an order rate and No Assets place an order the product of the difference of rate and lower single price of the user according to user, Determine the resource resources profit of the user.
- 5. according to the method described in claim 1, it is characterized in that, described have that resource places an order rate and No Assets place an order based on described Rate, before the step of determining the user corresponding resource resources profit, further includes:Pass through lower single price of lower single price estimation model pre-estimating user;It is described to have that resource places an order rate and No Assets place an order rate based on described, the step of determining the user corresponding resource resources profit, Including:Have that resource places an order rate and No Assets place an order rate according to user, and lower single price of the user and each resource Cost, determines the single resource resources profit of every kind of resource of the user;Using the maximum single resource resources profit as the corresponding resource resources profit of the user.
- 6. according to the method described in claim 5, it is characterized in that, described have resource to place an order rate and without money according to user Source places an order rate, and lower single price of the user and the cost of each resource, determines the single money of every kind of resource of the user The step of source income, including:Determine that the described of user has that resource places an order rate and No Assets place an order the difference of rate;According to the ratio of the difference and the product and the cost of single resource of lower single price of the user, the user is determined Single resource resources profit.
- 7. method according to any one of claims 1 to 6, it is characterised in that described pre- by the rate prediction model that places an order Estimate user have resource place an order rate and No Assets place an order rate the step of before, further include:Historical behavior data, user's representation data based on user, and user possess resource data, and the training rate of placing an order estimates mould Type;Wherein, the historical behavior data based on user, user's representation data, and user possess resource data, under training Single rate prediction model, including:According to the first user characteristic data and the first label data, the rate prediction model that places an order is trained;Wherein, first user characteristic data includes:The user behavior feature of default behavior, user's Figure Characteristics and for referring to Show that user possesses the resource characteristic of resource situation;Whether first label data places an order for instruction user;The default row For user behavior feature obtained according to the historical behavior data of the user, user's Figure Characteristics draw according to the user As data acquisition, the resource characteristic possesses resource data according to the user and obtains.
- 8. according to claim 3 to 6 any one of them method, it is characterised in that pass through lower single price estimation model described Before the step of estimating lower single price of user, further include:Historical behavior data and user's representation data based on the user for possessing resource, the lower single price estimation model of training;Wherein, The historical behavior data and user's representation data based on the user for possessing resource, the lower single price estimation model of training, including:According to the second user characteristic and the second label data of the user for possessing resource, the lower single price estimation model of training;Wherein, the second user characteristic includes:The corresponding price feature of default behavior, user's Figure Characteristics;Described Two label datas are used for lower single price of instruction user;The corresponding user characteristics of the default behavior is according to the history of the user Behavioral data obtains, and user's Figure Characteristics are obtained according to user's representation data.
- A kind of 9. resource allocation device, it is characterised in that including:The rate of placing an order estimates module, has that resource places an order rate and No Assets place an order for estimate user by the rate prediction model that places an order Rate;Resource resources profit determining module, for having of estimating that module estimates based on the rate of placing an order, resource places an order rate and No Assets place an order Rate, determines the corresponding resource resources profit of the user;Resource distribution module, it is right for the order of the resource resources profit that is determined according to the resource resources profit determining module from high to low Relative users carry out resource allocation.
- 10. a kind of electronic equipment, including memory, processor and it is stored on the memory and can runs on a processor Computer program, it is characterised in that the processor realizes claim 1 to 8 any one when performing the computer program The resource allocation methods.
- 11. a kind of computer-readable recording medium, is stored thereon with computer program, it is characterised in that the program is by processor The step of resource allocation methods described in claim 1 to 8 any one are realized during execution.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710942224.8A CN107944593A (en) | 2017-10-11 | 2017-10-11 | A kind of resource allocation methods and device, electronic equipment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710942224.8A CN107944593A (en) | 2017-10-11 | 2017-10-11 | A kind of resource allocation methods and device, electronic equipment |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107944593A true CN107944593A (en) | 2018-04-20 |
Family
ID=61936147
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710942224.8A Pending CN107944593A (en) | 2017-10-11 | 2017-10-11 | A kind of resource allocation methods and device, electronic equipment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107944593A (en) |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108876446A (en) * | 2018-06-01 | 2018-11-23 | 百度在线网络技术(北京)有限公司 | Consumption resources distribution method, device and equipment |
CN109191202A (en) * | 2018-08-28 | 2019-01-11 | 拉扎斯网络科技(上海)有限公司 | Resource allocation method, device, electronic equipment and computer readable storage medium |
CN109493113A (en) * | 2018-09-29 | 2019-03-19 | 口碑(上海)信息技术有限公司 | A kind of providing method and device of favor information |
CN109711887A (en) * | 2018-12-28 | 2019-05-03 | 拉扎斯网络科技(上海)有限公司 | Generation method and device of mall recommendation list, electronic equipment and computer medium |
CN109710410A (en) * | 2018-12-24 | 2019-05-03 | 微梦创科网络科技(中国)有限公司 | A kind of internet information resource distribution method and device |
CN109993428A (en) * | 2019-03-28 | 2019-07-09 | 第四范式(北京)技术有限公司 | Resource allocation methods and device |
CN110147882A (en) * | 2018-09-03 | 2019-08-20 | 腾讯科技(深圳)有限公司 | Training method, crowd's method of diffusion, device and the equipment of neural network model |
CN110400167A (en) * | 2019-06-28 | 2019-11-01 | 阿里巴巴集团控股有限公司 | User's screening technique and device |
CN110428278A (en) * | 2019-06-27 | 2019-11-08 | 阿里巴巴集团控股有限公司 | Determine the method and device of resource share |
CN110766430A (en) * | 2018-07-27 | 2020-02-07 | 深圳市立信创源科技有限公司 | Resource allocation method based on machine learning algorithm |
CN110858366A (en) * | 2018-08-24 | 2020-03-03 | 北京嘀嘀无限科技发展有限公司 | Method and device for predicting ordering conversion rate |
CN111639993A (en) * | 2020-05-29 | 2020-09-08 | 郑州轻工业大学 | Mobile data unloading and pricing method based on multi-item auction mechanism |
CN111985773A (en) * | 2020-07-15 | 2020-11-24 | 北京淇瑀信息科技有限公司 | User resource allocation strategy determining method and device and electronic equipment |
CN112073454A (en) * | 2020-07-22 | 2020-12-11 | 北京三快在线科技有限公司 | Resource distribution method and device and electronic equipment |
CN112381234A (en) * | 2020-11-09 | 2021-02-19 | 北京达佳互联信息技术有限公司 | Resource distribution method, device, equipment and computer readable storage medium |
CN113919868A (en) * | 2021-09-28 | 2022-01-11 | 上海画龙信息科技有限公司 | Equity distribution method and device based on marketing promotion model and electronic equipment |
CN113919869A (en) * | 2021-09-28 | 2022-01-11 | 上海画龙信息科技有限公司 | Equity distribution method and device based on sales increment model and electronic equipment |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106487715A (en) * | 2015-08-25 | 2017-03-08 | 百度在线网络技术(北京)有限公司 | A kind of method and apparatus for for user resource allocation |
CN106846071A (en) * | 2017-02-23 | 2017-06-13 | 北京小度信息科技有限公司 | Data processing method and device |
CN106846041A (en) * | 2016-12-26 | 2017-06-13 | 携程计算机技术(上海)有限公司 | The distribution method and system of reward voucher |
CN106910091A (en) * | 2017-02-23 | 2017-06-30 | 北京小度信息科技有限公司 | Data processing method and device |
-
2017
- 2017-10-11 CN CN201710942224.8A patent/CN107944593A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106487715A (en) * | 2015-08-25 | 2017-03-08 | 百度在线网络技术(北京)有限公司 | A kind of method and apparatus for for user resource allocation |
CN106846041A (en) * | 2016-12-26 | 2017-06-13 | 携程计算机技术(上海)有限公司 | The distribution method and system of reward voucher |
CN106846071A (en) * | 2017-02-23 | 2017-06-13 | 北京小度信息科技有限公司 | Data processing method and device |
CN106910091A (en) * | 2017-02-23 | 2017-06-30 | 北京小度信息科技有限公司 | Data processing method and device |
Cited By (26)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108876446A (en) * | 2018-06-01 | 2018-11-23 | 百度在线网络技术(北京)有限公司 | Consumption resources distribution method, device and equipment |
CN108876446B (en) * | 2018-06-01 | 2022-10-14 | 百度在线网络技术(北京)有限公司 | Consumption resource allocation method, device and equipment |
CN110766430A (en) * | 2018-07-27 | 2020-02-07 | 深圳市立信创源科技有限公司 | Resource allocation method based on machine learning algorithm |
CN110858366A (en) * | 2018-08-24 | 2020-03-03 | 北京嘀嘀无限科技发展有限公司 | Method and device for predicting ordering conversion rate |
CN109191202B (en) * | 2018-08-28 | 2021-09-07 | 拉扎斯网络科技(上海)有限公司 | Resource allocation method, device, electronic equipment and computer readable storage medium |
CN109191202A (en) * | 2018-08-28 | 2019-01-11 | 拉扎斯网络科技(上海)有限公司 | Resource allocation method, device, electronic equipment and computer readable storage medium |
CN110147882A (en) * | 2018-09-03 | 2019-08-20 | 腾讯科技(深圳)有限公司 | Training method, crowd's method of diffusion, device and the equipment of neural network model |
CN110147882B (en) * | 2018-09-03 | 2023-02-10 | 腾讯科技(深圳)有限公司 | Neural network model training method, crowd diffusion method, device and equipment |
CN109493113A (en) * | 2018-09-29 | 2019-03-19 | 口碑(上海)信息技术有限公司 | A kind of providing method and device of favor information |
CN109710410A (en) * | 2018-12-24 | 2019-05-03 | 微梦创科网络科技(中国)有限公司 | A kind of internet information resource distribution method and device |
CN109710410B (en) * | 2018-12-24 | 2021-08-13 | 微梦创科网络科技(中国)有限公司 | Internet information resource allocation method and device |
CN109711887B (en) * | 2018-12-28 | 2021-02-09 | 拉扎斯网络科技(上海)有限公司 | Generation method and device of mall recommendation list, electronic equipment and computer medium |
CN109711887A (en) * | 2018-12-28 | 2019-05-03 | 拉扎斯网络科技(上海)有限公司 | Generation method and device of mall recommendation list, electronic equipment and computer medium |
CN109993428A (en) * | 2019-03-28 | 2019-07-09 | 第四范式(北京)技术有限公司 | Resource allocation methods and device |
CN109993428B (en) * | 2019-03-28 | 2022-05-31 | 第四范式(北京)技术有限公司 | Resource allocation method and device |
CN110428278B (en) * | 2019-06-27 | 2023-10-27 | 创新先进技术有限公司 | Method and device for determining resource share |
CN110428278A (en) * | 2019-06-27 | 2019-11-08 | 阿里巴巴集团控股有限公司 | Determine the method and device of resource share |
CN110400167A (en) * | 2019-06-28 | 2019-11-01 | 阿里巴巴集团控股有限公司 | User's screening technique and device |
CN111639993B (en) * | 2020-05-29 | 2023-06-09 | 郑州轻工业大学 | Mobile data unloading and pricing method based on multi-item auction mechanism |
CN111639993A (en) * | 2020-05-29 | 2020-09-08 | 郑州轻工业大学 | Mobile data unloading and pricing method based on multi-item auction mechanism |
CN111985773A (en) * | 2020-07-15 | 2020-11-24 | 北京淇瑀信息科技有限公司 | User resource allocation strategy determining method and device and electronic equipment |
CN112073454A (en) * | 2020-07-22 | 2020-12-11 | 北京三快在线科技有限公司 | Resource distribution method and device and electronic equipment |
CN112381234A (en) * | 2020-11-09 | 2021-02-19 | 北京达佳互联信息技术有限公司 | Resource distribution method, device, equipment and computer readable storage medium |
CN112381234B (en) * | 2020-11-09 | 2024-05-14 | 北京达佳互联信息技术有限公司 | Resource distribution method, device, equipment and computer readable storage medium |
CN113919869A (en) * | 2021-09-28 | 2022-01-11 | 上海画龙信息科技有限公司 | Equity distribution method and device based on sales increment model and electronic equipment |
CN113919868A (en) * | 2021-09-28 | 2022-01-11 | 上海画龙信息科技有限公司 | Equity distribution method and device based on marketing promotion model and electronic equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107944593A (en) | A kind of resource allocation methods and device, electronic equipment | |
Farouk El Gayar et al. | An integrated framework for advanced hotel revenue management | |
CN110033314B (en) | Advertisement data processing method and device | |
Hu et al. | Using BPNN and DEMATEL to modify importance–performance analysis model–A study of the computer industry | |
Mahajan et al. | New product diffusion models in marketing: A review and directions for research | |
CN108154401A (en) | User's portrait depicting method, device, medium and computing device | |
CN103679504B (en) | Distribute the method and device of electronic bill | |
CN108133013A (en) | Information processing method, device, computer equipment and storage medium | |
CN108416614A (en) | Sending method, device, electronic equipment and the storage medium of virtual objects | |
CN101814170A (en) | Control device and control method for releasing advertisement in advertisement releasing and controlling system | |
CN113781139B (en) | Item recommending method, item recommending device, equipment and medium | |
CN108133390A (en) | For predicting the method and apparatus of user behavior and computing device | |
CN109727055A (en) | Network Questionnaire Survey method, apparatus, equipment and computer readable storage medium | |
CN107741986A (en) | User's behavior prediction and corresponding information recommend method and apparatus | |
CN109213936A (en) | product search method and device | |
Yu et al. | African entrepreneurs and international coordination in petty businesses: The case of low-end mobile phones sourcing in Hong Kong | |
CN118134553B (en) | E-commerce and explosion type multi-platform collaborative pushing system, method, equipment and medium | |
Yang et al. | Platform service offering to business customers: Strategic considerations in engendering seller use of marketing tools | |
CN108460611A (en) | A kind of information processing method and its application process and relevant device | |
CN112348356A (en) | Work quality determination method and device, computer equipment and readable storage medium | |
CN114625975B (en) | Knowledge graph-based customer behavior analysis system | |
Kaur et al. | An evaluation of middle-income group housing in Kolkata, India | |
CN108109042A (en) | Business favor information method for pushing and device | |
CN110264250B (en) | Method and device for determining distributed data of peer-to-peer resource quantity of product in multiple regions | |
CN107093103A (en) | Brand value evaluation method and system based on big data statistical analysis |
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 |