CN109165974A - A kind of commercial product recommending model training method, device, equipment and storage medium - Google Patents
A kind of commercial product recommending model training method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN109165974A CN109165974A CN201810884171.3A CN201810884171A CN109165974A CN 109165974 A CN109165974 A CN 109165974A CN 201810884171 A CN201810884171 A CN 201810884171A CN 109165974 A CN109165974 A CN 109165974A
- Authority
- CN
- China
- Prior art keywords
- time
- commodity
- user
- history
- variance
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0202—Market predictions or forecasting for commercial activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0203—Market surveys; Market polls
-
- 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]
- G06Q30/0631—Item recommendations
Abstract
The embodiment of the invention discloses a kind of commercial product recommending model training method, device, equipment and storage mediums, wherein, this method comprises: obtain each user browses the homepage of commodity, details page and the history residence time for evaluating page respectively, and obtain the history stop total time that each user browses commodity;The history residence time that each page is browsed according to each user determines the corresponding first time mean value of each page respectively, and determines first variance respectively;Total time is stopped according to the history, determines corresponding second time average of commodity, and determine second variance;According to first time mean value, first variance, the second time average and second variance, determine user to the interest-degree of commodity;Using the characteristic of commodity and the characteristic of user as input variable, setting model is trained using interest-degree as output variable, trained commercial product recommending model is obtained, it can be with Accurate Prediction user to the interest-degree of commodity, to increase the clicking rate of recommendation results.
Description
Technical field
The present embodiments relate to data processing technique more particularly to a kind of commercial product recommending model training method, device, set
Standby and storage medium.
Background technique
E-commerce platform is a big application field of individual commodity recommendation system.Wherein, individual commodity recommendation side
Formula includes: historical behavior data and real-time behavioral data based on user, recalls strategy by extensive stock and determines Recommendations
Candidate Set;The displaying sequence that Candidate Set commodity are determined by clicking rate prediction model is clicked the high commodity of probability and is preferentially opened up
Show.
Wherein, in the prior art, the modeling of clicking rate prediction model may is that exposure and click by user to commodity
Behavior constructs training set, and the label for exposing the sample not clicked on is 0, and the label for the sample for exposing and clicking is 1;Model is carried out
Training, then using each user of model prediction to the click probability of Candidate Set commodity, by clicking the determine the probability user's
Individual commodity recommendation sequence clicks the high commodity of probability and preferentially shows the user.
But the model of above-mentioned modeling cannot portray user to the interest-degree of commodity well.For example, user may be
The overdue commodity, the descriptive labelling or picture for being either only attracted people are confused, and leave this after clicking the commodity
Commodity related pages.Therefore, it is real that user's click probability that prediction model finally predicts can not be well reflected user
Interest.
Summary of the invention
The embodiment of the present invention provides a kind of commercial product recommending model training method, device, equipment and storage medium, can be accurate
User is predicted to the interest-degree of commodity, to increase the clicking rate of recommendation results.
In a first aspect, the embodiment of the invention provides a kind of commercial product recommending model training methods, comprising:
It obtains each user and browses the homepage of commodity, details page and the history residence time for evaluating page respectively, and obtain
The history that each user browses the commodity stops total time;
The history residence time that each page is browsed according to each user determines that each page corresponding first time is equal respectively
Value, and the corresponding first variance of the first time mean value is determined respectively;
Total time is stopped according to the history that each user browses the commodity, determines the commodity corresponding second
Time average, and determine the corresponding second variance of second time average;
According to the first time mean value, the first variance, second time average and the second variance, really
Interest-degree of the fixed user to the commodity;
It is defeated with the interest-degree using the characteristic of the commodity and the characteristic of the user as input variable
Variable is trained setting model out, obtains trained commercial product recommending model.
Second aspect, the embodiment of the invention also provides a kind of commercial product recommending model training apparatus, comprising:
Module is obtained, browses the homepage of commodity, details page and the history stop for evaluating page respectively for obtaining each user
Time, and obtain the history stop total time that each user browses the commodity;
First determining module determines each page for browsing the history residence time of each page according to each user respectively
The corresponding first time mean value in face, and the corresponding first variance of the first time mean value is determined respectively;
Second determining module, the history for browsing the commodity according to each user stop total time, determine
Corresponding second time average of the commodity, and determine the corresponding second variance of second time average;
Third determining module, for according to the first time mean value, the first variance, second time average with
And the second variance, determine the user to the interest-degree of the commodity;
Training module, for using the characteristic of the commodity and the characteristic of the user as input variable, with
The interest-degree is that output variable is trained setting model, obtains trained commercial product recommending model.
The third aspect, the embodiment of the invention also provides a kind of equipment, comprising:
One or more processors;
Storage device, for storing one or more programs,
When one or more of programs are executed by one or more of processors, so that one or more of processing
Device commercial product recommending model training method provided in an embodiment of the present invention.
Fourth aspect, the embodiment of the invention also provides a kind of computer readable storage mediums, are stored thereon with computer
Program, the program realize commercial product recommending model training method provided in an embodiment of the present invention when being executed by processor.
Technical solution provided in an embodiment of the present invention browses the history residence time of each page of commodity by each user, really
Determine the corresponding first time mean value of each page, and determine the corresponding first variance of first time mean value, passes through each user and browse quotient
The history of product stops total time, determines corresponding second time average of commodity, and determines the corresponding second party of the second time average
Difference determines user to the interest-degree of commodity by first time mean value, first variance, the second time average and second variance;
To carry out model training, training module is obtained, it can be with Accurate Prediction user to the interest-degree of commodity, to increase by the model
Add the clicking rate of recommendation results.
Detailed description of the invention
Fig. 1 is a kind of commercial product recommending model training method flow chart provided in an embodiment of the present invention;
Fig. 2 is a kind of commercial product recommending model training apparatus structural block diagram provided in an embodiment of the present invention;
Fig. 3 is a kind of device structure schematic diagram provided in an embodiment of the present invention.
Specific embodiment
The present invention is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is used only for explaining the present invention rather than limiting the invention.It also should be noted that in order to just
Only the parts related to the present invention are shown in description, attached drawing rather than entire infrastructure.
Fig. 1 is a kind of commercial product recommending model training method flow chart provided in an embodiment of the present invention, and the method is by one kind
Commercial product recommending model training apparatus executes, and described device is executed by software and/or hardware, described device can be integrated in end
In the equipment such as end, server.The method is applied in the scene recommended commodity.As shown in Figure 1, the embodiment of the present invention
The method of offer includes:
S110: obtaining each user and browse the homepage of commodity, details page and the history residence time for evaluating page respectively, and
Obtain the history stop total time that each user browses the commodity.
In embodiments of the present invention, user browse commodity history stop total time be user browse commodity homepage, in detail
Feelings page and the total time for evaluating page browse quotient for example, it is 20s that a user, which browses the history residence time of the homepage of commodity,
The history residence time of the details page of product is 50s, and the history residence time for browsing the evaluation page of commodity is 50s, then the user is clear
Look at commodity history stop total time be 50+50+20=120s.
S120: browsing the history residence time of each page according to each user, determines each page corresponding first respectively
Time average, and the corresponding first variance of the first time mean value is determined respectively.
S130: total time is stopped according to the history that each user browses the commodity, determines that the commodity are corresponding
The second time average, and determine the corresponding second variance of second time average.
In an embodiment of the embodiment of the present invention, when being stopped according to the history that each user browses each page
Between, the corresponding first time mean value of each page is determined respectively, and determines the corresponding first variance of the first time mean value respectively,
It include: logarithm to be taken to the history residence time of each page respectively, and calculate the average value for taking the history residence time of logarithm, and make
For the corresponding first time mean value of each page;According to the history residence time and the first time mean value for taking logarithm, determine
The first variance.
Correspondingly, the history stop for browsing the commodity according to each user determines the commodity total time
Corresponding second time average, and determine the corresponding second variance of second time average, comprising: the history is stopped total
Time takes logarithm, and calculates the average value for taking the history of logarithm to stop total time, and as the second time average;According to taking logarithm
History stop total time and second time average and determine the second variance.
Wherein, logarithm is taken to the history residence time of each page, then calculates average value, and as each page corresponding
One time average finally determines first variance.For example, 5 users are respectively to the history residence time of the homepage of a commodity
20s, 10s, 10s, 20s and 30s take logarithm to the history residence time respectively, that is, respectively correspond log20, log10, log10,
Then log20 and log30 calculates average value, i.e. (log20+log10+log10+log20+log30)/5, obtains average value, makees
For the corresponding first time mean value of homepage of the commodity.Then the calculation method of first variance can by log20, log10, log10,
Log20, log30 and (log20+log10+log10+log20+log30)/5 substitute into formula of variance and are calculated.
Wherein, the calculation method first time corresponding with homepage of the corresponding first time mean value of other pages of the commodity
The calculation method of mean value is identical, then the calculation method of first variance is also referred to above-mentioned method.And other commodity pages
The calculation method of corresponding first time mean value and first variance can refer to the above method.
Wherein, commodity corresponding second time average is determined total time according to the history stop that user browses commodity, thus
Determine the corresponding second variance of the second time average.For example, co-existing in 5 users, each user browses commodity history and stops always
Time is 100s, 120s, 110s, 50s and 30s, then takes logarithm total time to the stop of each history, then asked for the second time equal
Value, the second time average is log100+log120+log110+log50+log30)/5, then it can be calculated according to formula of variance
Obtain the corresponding second variance of the second time average.
Logarithm is taken by browsing the history residence time of each page to user as a result, and each quotient is browsed to user
The history stop of product takes logarithm total time, can reduce the numerical value of data, is convenient for data processing, facilitates calculating.
It is optionally, described that each page is browsed according to each user in an embodiment of the embodiment of the present invention
The history residence time determines the corresponding first time mean value of each page respectively, and determines that the first time mean value is corresponding respectively
First variance, can also include: the history residence time for browsing each page according to each user, calculate separately each page
The average value of history residence time, and as the corresponding first time mean value of each page;According to the history residence time and
The first time mean value, determines the first variance.
Correspondingly, the history for browsing the commodity according to each user stops total time, the quotient is determined
Corresponding second time average of product, and determine the corresponding second variance of second time average, it may include: according to each described
The history that user browses the commodity stops total time, and the history for calculating the commodity stops the average value of total time, and
As second time average;Total time is stopped according to history and second time average determines the second variance.
Wherein, logarithm can also not taken the history residence time to each page respectively, directly stopped by the history of each page
Time determines time average, so that it is determined that second variance.For example, when 5 users stop the history of the homepage of a commodity
Between be 20s, 10s, 10s, 20s and 30s respectively, then calculate the average value of the history residence time of homepage, and as at the first time
Mean value, as (20+10+10+20+30)/5=16, then the corresponding first variance of the first mean value is referred to formula of variance progress
It calculates, first variance can be calculated by average value and each history residence time.
Correspondingly, when calculating first time mean value and first variance, and the history residence time of each page is not sought
When logarithm, then the second time average and second variance are calculated, logarithm is not also sought to the history residence time of commodity.For example,
If co-existing in 5 users, each user's browsing commodity history stop total time is 100s, 120s, 110s, 50s and 30s, history
The average value for stopping total time is then (100+120+110+50+30)/5 as the second time average, and each history is stopped
Total time and the second time average, which are updated to variance calculation formula, can be obtained second variance.
S140: according to the first time mean value, the first variance, second time average and the second party
Difference determines the user to the interest-degree of the commodity.
In an embodiment of the embodiment of the present invention, optionally, it is described according to the first time mean value, described
One variance, the second time average and the second variance determine the user to the interest-degree of the commodity, comprising:
Determine the user to the interest-degree of the commodity according to following formula:
Wherein, PuiIt is u-th of user to the interest-degree of commodity i;Wherein, j is the number of the commodity page;
Wherein, tij=exp (μi+σ×zcj);μiTo take the history of logarithm to stop the average value of total time for commodity i,
Or being is corresponding second time average of commodity i, σiFor the μiCorresponding second variance;
Wherein, tcuijThe history residence time of the jth page of commodity i is browsed for u-th of user,
ucjFor the jth page for commodity i, the average value of the history residence time of logarithm is taken, or corresponding for the jth page of commodity i
First time mean value;σcjFor ucjCorresponding first variance.
Wherein, above-mentioned user is suitable for taking logarithm to the history residence time of each page to the calculation method of commercial productainterests degree
Determine first time mean value and first variance, and to the stop of the history of commodity take total time logarithm determine the second time average with
And the case where second variance.
It is optionally, described according to the first time mean value, institute in another embodiment of the embodiment of the present invention
First variance, second time average and the second variance are stated, determines the user to the interest-degree of the commodity, packet
It includes: determining the user to the interest-degree of the commodity according to following formula:
Wherein, PuiIt is u-th of user to the interest-degree of commodity i;Wherein, j is the number of the commodity page;
Wherein, tij=exp (μi+σ×zcj);μiThe average value or the commodity of total time are stopped for the history of commodity i
Corresponding second time average of i, σiFor the μiCorresponding second variance;
Wherein, tcuijThe history residence time of the jth page of commodity i is browsed for u-th of user;ucjFor
The average value of the history residence time of the jth page of commodity i, or it is equal for the corresponding first time of the jth page of commodity i
Value;σcjFor ucjCorresponding first variance.
Wherein, its above-mentioned user is suitable for the calculation method of commercial productainterests degree: passing through the history residence time of each page
Directly determine first time mean value and first variance, and it is equal by the stop of the history of commodity to directly determine for the second time total time
The case where value and second variance.
S150: using the characteristic of the commodity and the characteristic of the user as input variable, with the interest
Degree is that output variable is trained setting model, obtains trained commercial product recommending model.
Wherein, the characteristic of the user includes age or gender information of the user etc., product features data packet
Include Taxonomy Information, pricing information etc..The setting model is gradient boosted tree regression model.Wherein, setting model is not
It is confined to gradient boosted tree regression model, can also be other models.
In embodiments of the present invention, each user can be calculated to every by calculation formula of the user to commercial productainterests degree
The interest-degree of a commodity.The characteristic of user, the characteristic of commodity and user are input to the interest-degree of commodity and set
In cover half type, model is trained, adjusts the parameter in model.Wherein, when being trained to setting model, in order to make mould
Type prediction is accurate, and can choose multi-group data is trained setting model, the model after being trained.Wherein, multi-group data
Including multiple groups user characteristic data, product features data and multiple users are to the data of the interest-degree of commodity or other are more
The combination of group data.
On the basis of the above embodiments, further includes: by trained commercial product recommending model to the quotient in commodity Candidate Set
Product are predicted, the prediction interest-degree of each commodity is obtained, and are arranged according to prediction interest-degree the commodity in commodity Candidate Set
Sequence, and recommend user.Wherein, the commodity in commodity Candidate Set can be the commodity of user's history browsing, be also possible to basis
A kind of commodity for classification that user's operation determines, or can also be the commodity that other modes determine.
Technical solution provided in an embodiment of the present invention browses the history residence time of each page of commodity by each user, really
Determine the corresponding first time mean value of each page, and determine the corresponding first variance of first time mean value, passes through each user and browse quotient
The history of product stops total time, determines corresponding second time average of commodity, and determines the corresponding second party of the second time average
Difference, by first time mean value, first variance, the second time average and second variance, determine user to the interest-degree of commodity,
The duration of commodity is browsed by user to determine the interest-degree of user to user, to carry out model training, obtains training mould
Block, can be with Accurate Prediction user to the interest-degree of commodity, to increase the clicking rate of recommendation results by the model.
Fig. 2 is a kind of commercial product recommending model training apparatus structural block diagram provided in an embodiment of the present invention, as shown in Fig. 2, institute
Stating device includes obtaining module 210, the first determining module 220, the second determining module 230, third determining module 240 and training mould
Block 250.
Module 210 is obtained, the history for browsing the homepage of commodity, details page and evaluation page respectively for obtaining each user is stopped
The time is stayed, and obtains the history stop total time that each user browses the commodity;
First determining module 220 determines each respectively for browsing the history residence time of each page according to each user
The corresponding first time mean value of the page, and the corresponding first variance of the first time mean value is determined respectively;
Second determining module 230, the history for browsing the commodity according to each user stop total time, really
Determine corresponding second time average of the commodity, and determines the corresponding second variance of second time average;
Third determining module 240, for equal according to the first time mean value, the first variance, second time
Value and the second variance, determine the user to the interest-degree of the commodity;
Training module 250, for using the characteristic of the commodity and the characteristic of the user as input variable,
Setting model is trained using the interest-degree as output variable, obtains trained commercial product recommending model.
First determining module 220 takes logarithm for the history residence time respectively to each page, and calculates and take going through for logarithm
The average value of history residence time, and as the corresponding first time mean value of each page;
According to the history residence time and the first time mean value for taking logarithm, the first variance is determined;
Correspondingly, the second determining module 230, for taking logarithm total time to history stop, and calculates and takes logarithm
History stops the average value of total time, and as the second time average;
According to taking, the history of logarithm stops total time and second time average determines the second variance.
Optionally, third determining module 240, for determining the user to the interest of the commodity according to following formula
Degree:
Wherein, PuiIt is u-th of user to the interest-degree of commodity i;Wherein, j is the number of the commodity page;
Wherein, tij=exp (μi+σ×zcj);μiTo take the history of logarithm to stop the average value of total time for commodity i,
Or being is corresponding second time average of commodity i, σiFor the μiCorresponding second variance;
Wherein, tcuijThe history residence time of the jth page of commodity i is browsed for u-th of user,
ucjFor the jth page for commodity i, the average value of the history residence time of logarithm is taken, or corresponding for the jth page of commodity i
First time mean value;σcjFor ucjCorresponding first variance.
Optionally, the first determining module 220, for browsing the history residence time of each page according to each user, point
The average value of the history residence time of each page is not calculated, and as the corresponding first time mean value of each page;
According to the history residence time and the first time mean value, the first variance is determined;
Correspondingly, the second determining module 230, the history for browsing the commodity according to each user stops total
Time, the history for calculating the commodity stop the average value of total time, and as second time average;
Total time is stopped according to history and second time average determines the second variance.
Optionally, third determining module 240, for determining the user to the interest of the commodity according to following formula
Degree:
Wherein, PuiIt is u-th of user to the interest-degree of commodity i;Wherein, j is the number of the commodity page;
Wherein, tij=exp (μi+σ×zcj);μiThe average value or the commodity of total time are stopped for the history of commodity i
Corresponding second time average of i, σiFor the μiCorresponding second variance;
Wherein, tcuijThe history residence time of the jth page of commodity i is browsed for u-th of user;ucjFor
The average value of the history residence time of the jth page of commodity i, or it is equal for the corresponding first time of the jth page of commodity i
Value;σcjFor ucjCorresponding first variance.
Optionally, the setting model is gradient boosted tree regression model.
Optionally, the characteristic of the user includes age or the gender information of the user, product features data packet
Include Taxonomy Information.
Method provided by any embodiment of the invention can be performed in above-mentioned apparatus, has the corresponding functional module of execution method
And beneficial effect.
Fig. 3 is a kind of device structure schematic diagram provided in an embodiment of the present invention, as shown in figure 3, the equipment includes:
One or more processors 310, in Fig. 3 by taking a processor 310 as an example;
Memory 320;
The equipment can also include: input unit 330 and output device 340.
Processor 310, memory 320, input unit 330 and output device 340 in the equipment can pass through bus
Or other modes connect, in Fig. 3 for being connected by bus.
Memory 320 be used as a kind of non-transient computer readable storage medium, can be used for storing software program, computer can
Program and module are executed, such as the corresponding program instruction/mould of one of embodiment of the present invention commercial product recommending model training method
Block is (for example, attached acquisition module 210 shown in Fig. 2, the first determining module 220, the second determining module 230, third determining module
240 and training module 250).Software program, instruction and the module that processor 310 is stored in memory 320 by operation,
A kind of commodity thereby executing the various function application and data processing of computer equipment, i.e. realization above method embodiment push away
Recommend model training method, it may be assumed that
It obtains each user and browses the homepage of commodity, details page and the history residence time for evaluating page respectively, and obtain
The history that each user browses the commodity stops total time;
The history residence time that each page is browsed according to each user determines that each page corresponding first time is equal respectively
Value, and the corresponding first variance of the first time mean value is determined respectively;
Total time is stopped according to the history that each user browses the commodity, determines the commodity corresponding second
Time average, and determine the corresponding second variance of second time average;
According to the first time mean value, the first variance, second time average and the second variance, really
Interest-degree of the fixed user to the commodity;
It is defeated with the interest-degree using the characteristic of the commodity and the characteristic of the user as input variable
Variable is trained setting model out, obtains trained commercial product recommending model.
Memory 320 may include storing program area and storage data area, wherein storing program area can store operation system
Application program required for system, at least one function;Storage data area can be stored to be created according to using for computer equipment
Data etc..In addition, memory 320 may include high-speed random access memory, it can also include non-transitory memory, such as
At least one disk memory, flush memory device or other non-transitory solid-state memories.In some embodiments, it stores
Optional device 320 includes the memory remotely located relative to processor 310, these remote memories can be by being connected to the network extremely
Terminal device.The example of above-mentioned network includes but is not limited to internet, intranet, local area network, mobile radio communication and its group
It closes.
Input unit 330 can be used for receiving the number or character information of input, and generate the user with computer equipment
Setting and the related key signals input of function control.Output device 340 may include that display screen etc. shows equipment.
The embodiment of the invention provides a kind of computer readable storage mediums, are stored thereon with computer program, the program
Such as commercial product recommending model training method provided in an embodiment of the present invention is realized when being executed by processor:
It obtains each user and browses the homepage of commodity, details page and the history residence time for evaluating page respectively, and obtain
The history that each user browses the commodity stops total time;
The history residence time that each page is browsed according to each user determines that each page corresponding first time is equal respectively
Value, and the corresponding first variance of the first time mean value is determined respectively;
Total time is stopped according to the history that each user browses the commodity, determines the commodity corresponding second
Time average, and determine the corresponding second variance of second time average;
According to the first time mean value, the first variance, second time average and the second variance, really
Interest-degree of the fixed user to the commodity;
It is defeated with the interest-degree using the characteristic of the commodity and the characteristic of the user as input variable
Variable is trained setting model out, obtains trained commercial product recommending model.
It can be using any combination of one or more computer-readable media.Computer-readable medium can be calculating
Machine readable signal medium or computer readable storage medium.Computer readable storage medium for example can be --- but it is unlimited
In system, device or the device of --- electricity, magnetic, optical, electromagnetic, infrared ray or semiconductor, or any above combination.It calculates
The more specific example (non exhaustive list) of machine readable storage medium storing program for executing includes: electrical connection with one or more conducting wires, just
Taking formula computer disk, hard disk, random access memory (RAM), read-only memory (ROM), erasable type may be programmed read-only storage
Device (EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-ROM), light storage device, magnetic memory device,
Or above-mentioned any appropriate combination.In this document, computer readable storage medium can be it is any include or storage journey
The tangible medium of sequence, the program can be commanded execution system, device or device use or in connection.
Computer-readable signal media may include in a base band or as carrier wave a part propagate data-signal,
Wherein carry computer-readable program code.The data-signal of this propagation can take various forms, including --- but
It is not limited to --- electromagnetic signal, optical signal or above-mentioned any appropriate combination.Computer-readable signal media can also be
Any computer-readable medium other than computer readable storage medium, which can send, propagate or
Transmission is for by the use of instruction execution system, device or device or program in connection.
The program code for including on computer-readable medium can transmit with any suitable medium, including --- but it is unlimited
In --- wireless, electric wire, optical cable, RF etc. or above-mentioned any appropriate combination.
The computer for executing operation of the present invention can be write with one or more programming languages or combinations thereof
Program code, described program design language include object oriented program language-such as Java, Smalltalk, C++,
Further include conventional procedural programming language-such as " C " language or similar programming language.Program code can be with
It fully executes, partly execute on the user computer on the user computer, being executed as an independent software package, portion
Divide and partially executes or executed on a remote computer or server completely on the remote computer on the user computer.?
Be related in the situation of remote computer, remote computer can pass through the network of any kind --- including local area network (LAN) or
Wide area network (WAN)-be connected to subscriber computer, or, it may be connected to outer computer (such as mentioned using Internet service
It is connected for quotient by internet).
Note that the above is only a better embodiment of the present invention and the applied technical principle.It will be appreciated by those skilled in the art that
The invention is not limited to the specific embodiments described herein, be able to carry out for a person skilled in the art it is various it is apparent variation,
It readjusts and substitutes without departing from protection scope of the present invention.Therefore, although being carried out by above embodiments to the present invention
It is described in further detail, but the present invention is not limited to the above embodiments only, without departing from the inventive concept, also
It may include more other equivalent embodiments, and the scope of the invention is determined by the scope of the appended claims.
Claims (10)
1. a kind of commercial product recommending model training method characterized by comprising
It obtains each user and browses the homepage of commodity, details page and the history residence time for evaluating page respectively, and obtain each use
The history that family browses the commodity stops total time;
The history residence time that each page is browsed according to each user determines the corresponding first time mean value of each page respectively,
And the corresponding first variance of the first time mean value is determined respectively;
Total time is stopped according to the history that each user browses the commodity, determines the commodity corresponding second time
Mean value, and determine the corresponding second variance of second time average;
According to the first time mean value, the first variance, second time average and the second variance, institute is determined
User is stated to the interest-degree of the commodity;
It is that output becomes with the interest-degree using the characteristic of the commodity and the characteristic of the user as input variable
Amount is trained setting model, obtains trained commercial product recommending model.
2. the method according to claim 1, wherein the history for browsing each page according to each user is stopped
The time is stayed, determines the corresponding first time mean value of each page respectively, and determines the first time mean value corresponding first respectively
Variance, comprising:
Logarithm is taken to the history residence time of each page respectively, and calculates the average value for taking the history residence time of logarithm, and make
For the corresponding first time mean value of each page;
According to the history residence time and the first time mean value for taking logarithm, the first variance is determined;
Correspondingly, the history stop for browsing the commodity according to each user determines that the commodity are corresponding total time
The second time average, and determine the corresponding second variance of second time average, comprising:
Logarithm taken total time to history stop, and calculates the average value for taking the history of logarithm to stop total time, and as the
Two time averages;
According to taking, the history of logarithm stops total time and second time average determines the second variance.
3. according to the method described in claim 2, it is characterized in that, described according to the first time mean value, the first party
Difference, the second time average and the second variance, determine the user to the interest-degree of the commodity, comprising:
Determine the user to the interest-degree of the commodity according to following formula:
Wherein, PuiIt is u-th of user to the interest-degree of commodity i;Wherein, j is the number of the commodity page;
Wherein, tij=exp (μi+σ×zcj);μiFor commodity i, to take the history of logarithm to stop the average value of total time, or
To be corresponding second time average of commodity i, σiFor the μiCorresponding second variance;
Wherein, tcuijThe history residence time of the jth page of commodity i, u are browsed for u-th of usercjFor
For the jth page of commodity i, the average value of the history residence time of logarithm is taken, or is the jth page of commodity i corresponding
One time average;σcjFor ucjCorresponding first variance.
4. the method according to claim 1, wherein the history for browsing each page according to each user is stopped
The time is stayed, determines the corresponding first time mean value of each page respectively, and determines the first time mean value corresponding first respectively
Variance, comprising:
The history residence time that each page is browsed according to each user calculates separately being averaged for the history residence time of each page
Value, and as the corresponding first time mean value of each page;
According to the history residence time and the first time mean value, the first variance is determined;
Correspondingly, the history for browsing the commodity according to each user stops total time, the commodity pair are determined
The second time average answered, and determine the corresponding second variance of second time average, comprising:
According to each user browse the commodity the history stop total time, calculate the commodity history stop it is total when
Between average value, and as second time average;
Total time is stopped according to history and second time average determines the second variance.
5. according to the method described in claim 4, it is characterized in that, described according to the first time mean value, the first party
Poor, described second time average and the second variance determine the user to the interest-degree of the commodity, comprising:
Determine the user to the interest-degree of the commodity according to following formula:
Wherein, PuiIt is u-th of user to the interest-degree of commodity i;Wherein, j is the number of the commodity page;
Wherein, tij=exp (μi+σ×zcj);μiI pairs of the average value or the commodity of total time are stopped for the history of commodity i
The second time average answered, σiFor the μiCorresponding second variance;
Wherein, tcuijThe history residence time of the jth page of commodity i is browsed for u-th of user;ucjFor commodity
The average value of the history residence time of the jth page of i, or the corresponding first time mean value of the jth page for commodity i;σcj
For ucjCorresponding first variance.
6. the method according to claim 1, wherein the setting model is gradient boosted tree regression model.
7. the method according to claim 1, wherein the characteristic of the user includes the year of the user
Age or gender information, product features data include Taxonomy Information.
8. a kind of commercial product recommending model training apparatus characterized by comprising
Module is obtained, browses the homepage of commodity, details page and the history residence time for evaluating page respectively for obtaining each user,
And obtain the history stop total time that each user browses the commodity;
First determining module determines each page pair for browsing the history residence time of each page according to each user respectively
The first time mean value answered, and the corresponding first variance of the first time mean value is determined respectively;
Second determining module, the history for browsing the commodity according to each user stops total time, described in determination
Corresponding second time average of commodity, and determine the corresponding second variance of second time average;
Third determining module, for according to the first time mean value, the first variance, second time average and institute
Second variance is stated, determines the user to the interest-degree of the commodity;
Training module, for using the characteristic of the commodity and the characteristic of the user as input variable, with described
Interest-degree is that output variable is trained setting model, obtains trained commercial product recommending model.
9. a kind of equipment characterized by comprising
One or more processors;
Storage device, for storing one or more programs,
When one or more of programs are executed by one or more of processors, so that one or more of processors are real
Existing commercial product recommending model training method as claimed in claim 1.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is by processor
Commercial product recommending model training method as claimed in claim 1 is realized when execution.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810884171.3A CN109165974A (en) | 2018-08-06 | 2018-08-06 | A kind of commercial product recommending model training method, device, equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810884171.3A CN109165974A (en) | 2018-08-06 | 2018-08-06 | A kind of commercial product recommending model training method, device, equipment and storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109165974A true CN109165974A (en) | 2019-01-08 |
Family
ID=64898935
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810884171.3A Pending CN109165974A (en) | 2018-08-06 | 2018-08-06 | A kind of commercial product recommending model training method, device, equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109165974A (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110503482A (en) * | 2019-08-26 | 2019-11-26 | 深圳乐信软件技术有限公司 | A kind of articles handling method, device, terminal and storage medium |
CN111090756A (en) * | 2020-03-24 | 2020-05-01 | 腾讯科技(深圳)有限公司 | Artificial intelligence-based multi-target recommendation model training method and device |
CN111209386A (en) * | 2020-01-07 | 2020-05-29 | 重庆邮电大学 | Personalized text recommendation method based on deep learning |
CN111399947A (en) * | 2020-06-02 | 2020-07-10 | 平安国际智慧城市科技股份有限公司 | Application program guide page optimization pushing method and device and computer equipment |
CN112613951A (en) * | 2020-12-28 | 2021-04-06 | 赛尔网络有限公司 | Commodity page display method, device, equipment and medium |
CN113129053A (en) * | 2021-03-29 | 2021-07-16 | 北京沃东天骏信息技术有限公司 | Information recommendation model training method, information recommendation method and storage medium |
CN113268692A (en) * | 2021-05-18 | 2021-08-17 | 五八到家有限公司 | Method and system for automatically collecting customer options, electronic equipment and storage medium |
CN113706252A (en) * | 2021-08-30 | 2021-11-26 | 康键信息技术(深圳)有限公司 | Product recommendation method and device, electronic equipment and storage medium |
CN114385942A (en) * | 2022-01-14 | 2022-04-22 | 平安付科技服务有限公司 | Page element parameter optimization method, device and equipment based on user behavior |
CN117495513A (en) * | 2024-01-03 | 2024-02-02 | 深圳市智百威科技发展有限公司 | Shopping recommendation method and system based on big data and artificial intelligence |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104462156A (en) * | 2013-09-25 | 2015-03-25 | 阿里巴巴集团控股有限公司 | Feature extraction and individuation recommendation method and system based on user behaviors |
CN106021577A (en) * | 2016-06-01 | 2016-10-12 | 北京金山安全软件有限公司 | Information pushing method and device and electronic equipment |
CN106127594A (en) * | 2016-06-30 | 2016-11-16 | 武汉斗鱼网络科技有限公司 | User room interest-degree computational methods based on principal component analysis and system |
US20170186030A1 (en) * | 2015-04-21 | 2017-06-29 | Tencent Technology (Shenzhen) Company Limited | Advertisement click-through rate correction method and advertisement push server |
-
2018
- 2018-08-06 CN CN201810884171.3A patent/CN109165974A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104462156A (en) * | 2013-09-25 | 2015-03-25 | 阿里巴巴集团控股有限公司 | Feature extraction and individuation recommendation method and system based on user behaviors |
US20170186030A1 (en) * | 2015-04-21 | 2017-06-29 | Tencent Technology (Shenzhen) Company Limited | Advertisement click-through rate correction method and advertisement push server |
CN106021577A (en) * | 2016-06-01 | 2016-10-12 | 北京金山安全软件有限公司 | Information pushing method and device and electronic equipment |
CN106127594A (en) * | 2016-06-30 | 2016-11-16 | 武汉斗鱼网络科技有限公司 | User room interest-degree computational methods based on principal component analysis and system |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110503482A (en) * | 2019-08-26 | 2019-11-26 | 深圳乐信软件技术有限公司 | A kind of articles handling method, device, terminal and storage medium |
CN111209386A (en) * | 2020-01-07 | 2020-05-29 | 重庆邮电大学 | Personalized text recommendation method based on deep learning |
CN111209386B (en) * | 2020-01-07 | 2022-04-12 | 重庆邮电大学 | Personalized text recommendation method based on deep learning |
CN111090756A (en) * | 2020-03-24 | 2020-05-01 | 腾讯科技(深圳)有限公司 | Artificial intelligence-based multi-target recommendation model training method and device |
CN111399947A (en) * | 2020-06-02 | 2020-07-10 | 平安国际智慧城市科技股份有限公司 | Application program guide page optimization pushing method and device and computer equipment |
CN112613951A (en) * | 2020-12-28 | 2021-04-06 | 赛尔网络有限公司 | Commodity page display method, device, equipment and medium |
CN113129053A (en) * | 2021-03-29 | 2021-07-16 | 北京沃东天骏信息技术有限公司 | Information recommendation model training method, information recommendation method and storage medium |
CN113268692A (en) * | 2021-05-18 | 2021-08-17 | 五八到家有限公司 | Method and system for automatically collecting customer options, electronic equipment and storage medium |
CN113706252A (en) * | 2021-08-30 | 2021-11-26 | 康键信息技术(深圳)有限公司 | Product recommendation method and device, electronic equipment and storage medium |
CN114385942A (en) * | 2022-01-14 | 2022-04-22 | 平安付科技服务有限公司 | Page element parameter optimization method, device and equipment based on user behavior |
CN117495513A (en) * | 2024-01-03 | 2024-02-02 | 深圳市智百威科技发展有限公司 | Shopping recommendation method and system based on big data and artificial intelligence |
CN117495513B (en) * | 2024-01-03 | 2024-03-29 | 深圳市智百威科技发展有限公司 | Shopping recommendation method and system based on big data and artificial intelligence |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109165974A (en) | A kind of commercial product recommending model training method, device, equipment and storage medium | |
CN109104620B (en) | Short video recommendation method and device and readable medium | |
US10497013B2 (en) | Purchasing behavior analysis apparatus and non-transitory computer readable medium | |
US11521221B2 (en) | Predictive modeling with entity representations computed from neural network models simultaneously trained on multiple tasks | |
CN112000819B (en) | Multimedia resource recommendation method and device, electronic equipment and storage medium | |
CN109299327A (en) | Video recommendation method, device, equipment and storage medium | |
CN101517967B (en) | Traffic prediction for web sites | |
CN109460512B (en) | Recommendation information processing method, device, equipment and storage medium | |
CN109460514A (en) | Method and apparatus for pushed information | |
CN107908789A (en) | Method and apparatus for generating information | |
EP4181026A1 (en) | Recommendation model training method and apparatus, recommendation method and apparatus, and computer-readable medium | |
US11172040B2 (en) | Method and apparatus for pushing information | |
CN110473042B (en) | Method and device for acquiring information | |
CN114417174B (en) | Content recommendation method, device, equipment and computer storage medium | |
CN107977678A (en) | Method and apparatus for output information | |
CN112182281B (en) | Audio recommendation method, device and storage medium | |
US11962662B2 (en) | Method and apparatus for pushing information | |
CN110489531A (en) | The determination method and apparatus of high frequency problem | |
CN110717599B (en) | Dissociation characterization learning method and device integrating multiple modes | |
CN112785391B (en) | Recommendation processing method and device, intelligent equipment and storage medium | |
CN111507471B (en) | Model training method, device, equipment and storage medium | |
CN111125501B (en) | Method and device for processing information | |
CN113743973A (en) | Method and device for analyzing market hotspot trend | |
CN111199454B (en) | Real-time user conversion evaluation method and device and electronic equipment | |
CN116911912B (en) | Method and device for predicting interaction objects and interaction results |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190108 |
|
RJ01 | Rejection of invention patent application after publication |