CN109785072A - Method and apparatus for generating information - Google Patents

Abstract

The embodiment of the present application discloses the method and apparatus for generating information.One specific embodiment of this method includes: the attribute information for obtaining target product；Based on attribute information and preset algorithm set, candidate value score set is determined；By candidate value score set input order model trained in advance, probability value set is obtained, wherein the probability value in probability value set corresponds to the algorithm in algorithm set, for characterizing the order of accuarcy of the candidate value score determined using corresponding algorithm；The selection target probability value from probability value set, and candidate label and the storage for being worth score and being determined as target product that will be determined according to the corresponding algorithm of destination probability value.This embodiment improves the accuracy for the label for generating target product, facilitate the specific aim that recommendation information is improved using the label generated.

Description

Method and apparatus for generating information

Technical field

The invention relates to field of computer technology, and in particular to the method and apparatus for generating information.

Background technique

Currently, in order to user's more targetedly recommendation information, it usually needs advance with the user information of user with And the product information of various products classifies to user and product, under the user to establish under each classification and each classification The corresponding relationship of product.It may include being drawn in advance to the property of value value of various products to the method that product is classified Point, multiple property of value values section is obtained, the product that property of value value is in some section is divided into a kind of product.

Summary of the invention

The embodiment of the present application proposes the method and apparatus for generating information, and method and dress for pushed information It sets.

In a first aspect, the embodiment of the present application provides a kind of method for generating information, this method comprises: obtaining target The attribute information of product；Based on attribute information and preset algorithm set, candidate value score set is determined；By candidate's value point Manifold closes input order model trained in advance, obtains probability value set, wherein the probability value in probability value set, which corresponds to, to be calculated Algorithm in method set, for characterizing the order of accuarcy of the candidate value score determined using corresponding algorithm；From probability value collection Selection target probability value in conjunction, and the candidate value score determined according to the corresponding algorithm of destination probability value is determined as target The label of product and storage.

In some embodiments, attribute information includes the order volume of target product in the target time period；And based on category Property information and preset algorithm set, determine candidate value score set, comprising: determine target product in the target time period Whether order volume is more than or equal to preset order volume threshold value；It is more than or equal to order volume threshold value in response to determining, is based on attribute information With the algorithm in preset algorithm set, determine candidate value score set as the first candidate value score set；In response to It determines and is less than order volume threshold value, based on other algorithms in attribute information and preset algorithm set, in addition to target algorithm, Determine candidate value score set as the second candidate value score set, wherein target algorithm be in the target time period Order volume be more than or equal to order volume threshold value the corresponding algorithm of product.

In some embodiments, target algorithm includes: in response to determining the order volume of target product in the target time period More than or equal to preset order volume threshold value, product that is predetermined, obtaining user in the user group of target product is obtained Obtain correlation；Correlation is obtained based on obtained product, determines the candidate value score of target product.

In some embodiments, target algorithm includes: in response to determining the order volume of target product in the target time period More than or equal to preset order volume threshold value, from preset at least two value score section, at least one following value is determined Score section: the corresponding original value attribute value for being worth score section, target product of the real value attribute value of target product Corresponding value score section；For the value score section at least one identified value score section, obtain default , corresponding with value score section transformation coefficient, and determined from the candidate value score of fixed target product Target candidate is worth score；It is worth score based on identified transformation coefficient and target candidate, determines the new time of target product Choosing value score.

In some embodiments, order model includes the first model and the second model；And

By candidate value score set input order model trained in advance, probability value set is obtained, comprising: in response to true The order volume of product in the target time period that sets the goal is more than or equal to order volume threshold value, by the first candidate value score set input First model obtains at least two probability values；In response to determining that the order volume of target product in the target time period is less than order Threshold value is measured, the second candidate value score set is inputted into the second model, obtains at least two probability values.

In some embodiments, attribute information comprises at least one of the following numerical value: the real value attribute value of target product, The original value attribute value of target product；And algorithm set comprises at least one of the following algorithm: based on belonging to target product The real value attribute value for the product that classification includes, arranges the product information for the product that classification belonging to target product includes Sequence obtains at least two first product information sequences, for first in obtained at least two first product information sequence Product information sequence, the arrangement position based on target product in the first product information sequence, determines that target product corresponds to The candidate value score of the first product information sequence；Original value category based on the product that classification belonging to target product includes Property value, the product information for the product that classification belonging to target product includes is ranked up, obtain at least two second products letter Sequence is ceased, for the second product information sequence in obtained at least two second product information sequence, is based on target product Arrangement position in the second product information sequence determines that target product corresponds to the candidate valence of the second product information sequence It is worth score.

In some embodiments, algorithm set further includes following at least one algorithm: determining classification belonging to target product Including product real value attribute value mean value and standard deviation；Based on identified mean value and standard deviation, determine that target produces The candidate value score of product；Determine the mean value and standard of the original value attribute value for the product that classification belonging to target product includes Difference；Based on identified mean value and standard deviation, the candidate value score of target product is determined.

In some embodiments, training obtains order model in accordance with the following steps in advance: sample attribute information aggregate is obtained, Wherein, sample attribute information corresponds to sample product；For the sample attribute information in sample attribute information aggregate, it is based on the sample This attribute information determines that the corresponding sample candidate of the sample attribute information is worth score set using algorithm set；Based on gained The sample candidate arrived is worth score set and value score section preset, corresponding with the sample attribute information, generates at least One markup information, wherein markup information corresponds to the algorithm in algorithm set, and markup information utilizes corresponding calculation for characterizing The sample candidate that method determines is worth whether score is located at the corresponding value score section of sample product；Using machine learning method, It, will be with input using the corresponding sample candidate value score set of the sample attribute information in sample attribute information aggregate as input Sample candidate be worth the corresponding markup information of score set and be used as desired output, train and obtain order model.

In some embodiments, the corresponding algorithm of markup information at least one markup information generated is preparatory It determines in accordance with the following steps: for the algorithm in algorithm set, determining that the sample candidate according to determined by the algorithm is worth score Accuracy rate；Based on identified accuracy rate, selection algorithm is as algorithm corresponding with markup information from algorithm set.

In some embodiments, order model is the multi-tag disaggregated model obtained based on Random Forest model training.

Second aspect, the embodiment of the present application provide a kind of method for pushed information, this method comprises: obtaining target The classification information of user and preset product information set, wherein product information corresponds to predetermined label, and label is root It is generated according to the method that any embodiment in above-mentioned first aspect describes；Correspondence based on preset, label and classification information Relationship, from product information set, the determining and matched product information of classification information；Identified product information is pushed into mesh Mark the terminal of user.

The third aspect, the embodiment of the present application provide it is a kind of for generating the device of information, the device include: obtain it is single Member is configured to obtain the attribute information of target product；Determination unit is configured to based on attribute information and preset set of algorithms It closes, determines candidate value score set；Generation unit is configured to the order that candidate value score set input is trained in advance Model obtains probability value set, wherein the probability value in probability value set corresponds to the algorithm in algorithm set, for characterizing Utilize the order of accuarcy for the candidate value score that corresponding algorithm determines；Storage unit is configured to select from probability value set Destination probability value is selected, and the candidate value score determined according to the corresponding algorithm of destination probability value is determined as target product Label and storage.

In some embodiments, attribute information includes the order volume of target product in the target time period；And it determines single Member includes: the first determining module, be configured to determine the order volume of target product in the target time period whether be more than or equal to it is pre- If order volume threshold value；Second determining module is configured in response to determine more than or equal to order volume threshold value, is based on attribute information With the algorithm in preset algorithm set, determine candidate value score set as the first candidate value score set；Second really Cover half block, be configured in response to determine be less than order volume threshold value, based on it is in attribute information and preset algorithm set, remove Other algorithms other than target algorithm determine candidate value score set as the second candidate value score set, wherein target Algorithm is algorithm corresponding more than or equal to the product of order volume threshold value with order volume in the target time period.

In some embodiments, target algorithm includes: in response to determining the order volume of target product in the target time period More than or equal to preset order volume threshold value, product that is predetermined, obtaining user in the user group of target product is obtained Obtain correlation；Correlation is obtained based on obtained product, determines the candidate value score of target product.

In some embodiments, target algorithm includes: in response to determining the order volume of target product in the target time period More than or equal to preset order volume threshold value, from preset at least two value score section, at least one following value is determined Score section: the corresponding original value attribute value for being worth score section, target product of the real value attribute value of target product Corresponding value score section；For the value score section at least one identified value score section, obtain default , corresponding with value score section transformation coefficient, and determined from the candidate value score of fixed target product Target candidate is worth score；It is worth score based on identified transformation coefficient and target candidate, determines the new time of target product Choosing value score.

In some embodiments, order model includes the first model and the second model；And generation unit includes: first raw At module, it is configured in response to determine that the order volume of target product in the target time period is more than or equal to order volume threshold value, it will First candidate value score set inputs the first model, obtains at least two probability values；Second generation module, is configured to respond to In determining that the order volume of target product in the target time period is less than order volume threshold value, by the second candidate value score set input Second model obtains at least two probability values.

In some embodiments, attribute information comprises at least one of the following numerical value: the real value attribute value of target product, The original value attribute value of target product；And algorithm set comprises at least one of the following algorithm: based on belonging to target product The real value attribute value for the product that classification includes, arranges the product information for the product that classification belonging to target product includes Sequence obtains at least two first product information sequences, for first in obtained at least two first product information sequence Product information sequence, the arrangement position based on target product in the first product information sequence, determines that target product corresponds to The candidate value score of the first product information sequence；Original value category based on the product that classification belonging to target product includes Property value, the product information for the product that classification belonging to target product includes is ranked up, obtain at least two second products letter Sequence is ceased, for the second product information sequence in obtained at least two second product information sequence, is based on target product Arrangement position in the second product information sequence determines that target product corresponds to the candidate valence of the second product information sequence It is worth score.

In some embodiments, algorithm set further includes following at least one algorithm: determining classification belonging to target product Including product real value attribute value mean value and standard deviation；Based on identified mean value and standard deviation, determine that target produces The candidate value score of product；Determine the mean value and standard of the original value attribute value for the product that classification belonging to target product includes Difference；Based on identified mean value and standard deviation, the candidate value score of target product is determined.

In some embodiments, training obtains order model in accordance with the following steps in advance: sample attribute information aggregate is obtained, Wherein, sample attribute information corresponds to sample product；For the sample attribute information in sample attribute information aggregate, it is based on the sample This attribute information determines that the corresponding sample candidate of the sample attribute information is worth score set using algorithm set；Based on gained The sample candidate arrived is worth score set and value score section preset, corresponding with the sample attribute information, generates at least One markup information, wherein markup information corresponds to the algorithm in algorithm set, and markup information utilizes corresponding calculation for characterizing The sample candidate that method determines is worth whether score is located at the corresponding value score section of sample product；Using machine learning method, It, will be with input using the corresponding sample candidate value score set of the sample attribute information in sample attribute information aggregate as input Sample candidate be worth the corresponding markup information of score set and be used as desired output, train and obtain order model.

In some embodiments, the corresponding algorithm of markup information at least one markup information generated is preparatory It determines in accordance with the following steps: for the algorithm in algorithm set, determining that the sample candidate according to determined by the algorithm is worth score Accuracy rate；Based on identified accuracy rate, selection algorithm is as algorithm corresponding with markup information from algorithm set.

In some embodiments, order model is the multi-tag disaggregated model obtained based on Random Forest model training.

Fourth aspect, the embodiment of the present application provide a kind of device for pushed information, which includes: to obtain list Member is configured to obtain the classification information and preset product information set of target user, wherein product information corresponds to preparatory Determining label, label are that the method described according to any embodiment in above-mentioned first aspect generates；Determination unit is matched It is set to the corresponding relationship based on preset, label and classification information, from product information set, determination is matched with classification information Product information；Push unit is configured to push to identified product information the terminal of target user.

5th aspect, the embodiment of the present application provide a kind of server, which includes: one or more processors； Storage device is stored thereon with one or more programs；When one or more programs are executed by one or more processors, so that One or more processors realize the method as described in implementation any in first aspect or second aspect.

6th aspect, the embodiment of the present application provide a kind of computer-readable medium, are stored thereon with computer program, should The method as described in implementation any in first aspect or second aspect is realized when computer program is executed by processor.

Method and apparatus provided by the embodiments of the present application for generating information, pass through the attribute information based on target product With preset algorithm set, candidate value score set, then the order that candidate value score set input is trained in advance are determined Model obtains probability value set, finally the selection target probability value from probability value set, and will be corresponding according to destination probability value The candidate value score that determines of algorithm be determined as label and the storage of target product, thus using candidate value score set and Order model improves the accuracy for generating the label of target product, helps to improve recommendation information using the label generated Specific aim.

Detailed description of the invention

By reading a detailed description of non-restrictive embodiments in the light of the attached drawings below, the application's is other Feature, objects and advantages will become more apparent upon:

Fig. 1 is that one embodiment of the application can be applied to exemplary system architecture figure therein；

Fig. 2 is the flow chart according to one embodiment of the method for generating information of the embodiment of the present application；

Fig. 3 is the schematic diagram according to an application scenarios of the method for generating information of the embodiment of the present application；

Fig. 4 is the flow chart according to another embodiment of the method for generating information of the embodiment of the present application；

Fig. 5 is the flow chart according to one embodiment of the method for pushed information of the embodiment of the present application；

Fig. 6 is the structural schematic diagram according to one embodiment of the device for generating information of the embodiment of the present application；

Fig. 7 is the structural schematic diagram according to one embodiment of the device for pushed information of the embodiment of the present application；

Fig. 8 is adapted for the structural schematic diagram for the computer system for realizing the server of the embodiment of the present application.

Specific embodiment

The application is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched The specific embodiment stated is used only for explaining related invention, rather than the restriction to the invention.It also should be noted that in order to Convenient for description, part relevant to related invention is illustrated only in attached drawing.

It should be noted that in the absence of conflict, the features in the embodiments and the embodiments of the present application can phase Mutually combination.The application is described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

Fig. 1 is shown can the method for generating information using the embodiment of the present application or the dress for generating information It sets, and the method for pushed information or the exemplary system architecture of the device for pushed information 100.

As shown in Figure 1, system architecture 100 may include terminal device 101,102,103, network 104 and server 105. Network 104 between terminal device 101,102,103 and server 105 to provide the medium of communication link.Network 104 can be with Including various connection types, such as wired, wireless communication link or fiber optic cables etc..

User can be used terminal device 101,102,103 and be interacted by network 104 with server 105, to receive or send out Send message etc..Various telecommunication customer end applications, such as the application of shopping class, net can be installed on terminal device 101,102,103 The application of page browsing device, searching class application, instant messaging tools, mailbox client, social platform software etc..

Terminal device 101,102,103 can be hardware, be also possible to software.When terminal device 101,102,103 is hard When part, it can be various electronic equipments, including but not limited to smart phone, tablet computer, pocket computer on knee and desk-top Computer etc..When terminal device 101,102,103 is software, may be mounted in above-mentioned cited electronic equipment.Its Multiple softwares or software module (such as providing the software of Distributed Services or software module) may be implemented into, it can also be real Ready-made single software or software module.It is not specifically limited herein.

Server 105 can be to provide the server of various services, such as obtain to from terminal device 101,102,103 The background information processing server that is handled of attribute information.Background information processing server can be to attribute information at Reason, and generate processing result (such as label of target product) and storage.

It should be noted that the side provided by the embodiment of the present application for generating the method for information or for pushed information Method is generally executed by server 105, and correspondingly, the device for generating the device of information or for pushed information is generally positioned at In server 105.

It should be noted that server can be hardware, it is also possible to software.When server is hardware, may be implemented At the distributed server cluster that multiple servers form, individual server also may be implemented into.It, can when server is software To be implemented as multiple softwares or software module (such as providing the software of Distributed Services or software module), also may be implemented At single software or software module.It is not specifically limited herein.

It should be understood that the number of terminal device, network and server in Fig. 1 is only schematical.According to realization need It wants, can have any number of terminal device, network and server.

With continued reference to Fig. 2, the process of one embodiment of the method for generating information according to the application is shown 200.The method for being used to generate information, comprising the following steps:

Step 201, the attribute information of target product is obtained.

In the present embodiment, can lead to for generating the executing subject (such as server shown in FIG. 1) of the method for information Wired connection mode or radio connection are crossed from attribute information long-range or from local acquisition target product.Wherein, target Product can be the product of its label to be determined, such as can be the product of the product information characterization of certain website offer.Attribute letter Breath may include information relevant to each attribute of target product.For example, attribute information may include following at least one letter Breath: title, weight, size, the property of value value of the property of value for characterizing target product of target product etc..

Step 202, it is based on attribute information and preset algorithm set, determines candidate value score set.

In the present embodiment, based on the attribute information and preset algorithm set obtained in step 201, above-mentioned executing subject It can determine candidate value score set.Wherein, candidate value score can be used for characterizing the size of the value of target product.

Specifically, attribute information includes the numerical value for characterizing the property of value of target product, which can be used for table Requisition family obtains the cost that target product is paid.For the algorithm in above-mentioned algorithm set, the algorithm and above-mentioned category are utilized Property the information property of value for characterizing target product that includes numerical value, the corresponding candidate value of the algorithm can be calculated Score.In general, the algorithm in algorithm set can be characterized in the form of formula or application program, the value of target product will be characterized The numerical value of attribute is input to formula or application program, available candidate value score.

In some optional implementations of the present embodiment, above-mentioned attribute information comprises at least one of the following numerical value: mesh Mark the real value attribute value of product, the original value attribute value of target product.In general, real value attribute value is in original valence The numerical value of certain numerical value is reduced on the basis of value attribute value, user is practical to get target production to real value attribute value for characterizing The cost that product are paid.Above-mentioned algorithm set may include following at least one algorithm:

Algorithm one, firstly, the real value attribute value for the product for including based on classification belonging to target product, produces target The product information for the product that classification belonging to product includes is ranked up, and obtains at least two first product information sequences.Wherein, it produces Product information can be the information for characterizing product, such as the number of product.The attribute information of product may include belonging to product Classification, above-mentioned executing subject can according to preset product concentrate product attribute information, determine belonging to target product The product that classification includes.Then, for the first product information sequence in obtained at least two first product information sequence, Arrangement position based on target product in the first product information sequence determines that target product corresponds to first product information The candidate value score of sequence.

Specifically, as an example, to Mr. Yu's the first product information sequence, which can be according to as follows Mode one obtains: according to the corresponding real value attribute value of product information it is descending be ranked up after obtain the first product letter Cease sequence.Each product information that the first product information sequence includes corresponds to a product, i.e., each product information is different.

Alternatively, the first product information sequence can be obtained with as follows two: acquisition target time section (such as recently The periods such as one day, one week) in the order of each product concentrated of preset product, to the corresponding product of each order of acquisition Product information, obtain the first product information after being ranked up according to the corresponding real value attribute value of product information is descending Sequence.That is, each product information that the first product information sequence includes corresponds to the order of a product, first product letter Breath sequence includes multiple subsequences being made of identical product information, and each subsequence corresponds to a product.

For aforesaid way one, above-mentioned executing subject can determine the candidate value score of target product are as follows: 1-k/N.Its In, k is serial number of the product information of target product in the product information sequence, and N is the product that the product information sequence includes The sum of information.For mode two, above-mentioned executing subject can determine the corresponding candidate value score undetermined of some product information Are as follows: 1-k/N, wherein k is serial number of the product information in the product information sequence, and N is the production that the product information sequence includes The sum of product information.Then, above-mentioned executing subject can determine that the product information that the corresponding subsequence of target product includes is corresponding The mean value of candidate value score undetermined be worth score as the candidate of target product.

As another example, to Mr. Yu's the first product information sequence, which can be according to such as lower section Formula three obtains: to the corresponding real value attribute value of product information carry out Logarithmic calculation, according to the numerical value being calculated by greatly to Small sequence, the first product information sequence obtained after being ranked up to each product information.

Alternatively, the first product information sequence can be obtained with as follows four: acquisition target time section (such as recently The periods such as one day, one week) in the order of each product concentrated of preset product, to the corresponding product of each order of acquisition Real value attribute value carry out Logarithmic calculation, according to the sequence that the numerical value being calculated is descending, to each product information The the first product information sequence obtained after being ranked up.That is, each product information that the first product information sequence includes is corresponding In the order of a product, which includes multiple subsequences being made of identical product information, each Subsequence corresponds to a product.Wherein, the formula of Logarithmic calculation can be with are as follows: logp_i, wherein the bottom of the logarithm is that 10, i is Serial number of the product information in the first product information queue, p_iFor the real value attribute value of i-th of product.

For aforesaid way three, above-mentioned executing subject can determine the candidate value score of target product are as follows: 1-k/N.Its In, k is serial number of the product information of target product in the product information sequence, and N is the product that the product information sequence includes The sum of information.For aforesaid way four, above-mentioned executing subject can determine the corresponding candidate value undetermined of some product information Score are as follows: 1-k/N, wherein k is serial number of the product information in the product information sequence, and N is that the product information sequence includes Product information sum.Then, above-mentioned executing subject can determine the product information that the corresponding subsequence of target product includes Candidate value score of the mean value of corresponding candidate value score undetermined as target product.

Algorithm two, the original value attribute value based on the product that classification belonging to target product includes, to target product institute The product information for the product that the classification of category includes is ranked up, and at least two second product information sequences is obtained, for acquired At least two second product information sequences in the second product information sequence, based on target product in the second product information sequence Arrangement position in column determines that target product corresponds to the candidate value score of the second product information sequence.

Specifically, for algorithm two, (algorithm one can be directed to by above-mentioned to the similar method of algorithm one according to above-mentioned Citing in real value attribute value replace with original value attribute value) determine target product correspond to the second product information sequence The candidate value score of column.

In some optional implementations of the present embodiment, above-mentioned algorithm set can also comprise at least one of the following calculation Method:

Algorithm three determines the mean value and standard of the real value attribute value for the product that classification belonging to target product includes Difference；Based on identified mean value and standard deviation, the candidate value score of target product is determined.

Specifically, it calculates mean value and the formula of standard deviation is as follows:

Wherein, μ is mean value, and σ is standard deviation, and N is the sum for the product information that the product information sequence includes, and i is target The serial number of the position of the characterization for the product that classification belonging to product includes in the sequence, p_iFor the real value category of i-th of product Property value, the bottom of the logarithm in the formula is 10.

Then, it is based on above-mentioned mean value and standard deviation, the candidate value of target product can be calculated according to following formula Score:

Wherein, q is candidate value score, and p is the real value attribute value of target product, and α, β are respectively preset numerical value (such as α=3, β=2).The bottom of logarithm in the formula is 10.

In addition, being based on above-mentioned mean value and standard deviation, the candidate valence of target product can also be calculated according to following formula It is worth score:

Wherein, e is natural constant, [e^μ-βσ,e^μ+ασ] be real value attribute value section.It, can be by mesh by the formula The real value attribute value of mark product is mapped to [0,1] section, and (can be greater than e for larger^μ+ασ) real value attribute It is worth and smaller (i.e. less than e^μ-βσ) real value attribute value be respectively set to 1 and 0.To reach rejecting abnormalities real value category The purpose of property value.

Algorithm four determines the mean value and standard of the original value attribute value for the product that classification belonging to target product includes Difference；Based on identified mean value and standard deviation, the candidate value score of target product is determined.

Specifically, for algorithm four, (algorithm three can be directed to by above-mentioned to the similar method of algorithm three according to above-mentioned Citing in real value attribute value replace with original value attribute value) determine the candidate value score of target product.

Step 203, the order model that candidate value score set input is trained in advance, obtains probability value set.

In the present embodiment, the order mould that above-mentioned executing subject can be trained in advance by candidate value score set input Type obtains probability value set.Wherein, the probability value in probability value set corresponds to the algorithm in algorithm set, for characterizing benefit With the order of accuarcy for the candidate value score that corresponding algorithm determines.

In general, candidate value score set can input order model, the probability value set of output in vector form It can be the form of vector.The quantity for the probability value that probability value set includes is less than the candidate valence that candidate value score set includes It is worth the quantity of score.Each probability value corresponds to an algorithm in above-mentioned algorithm set.Probability value is higher, indicates general according to this The accuracy that rate is worth the candidate value score that corresponding algorithm determines is higher.

The above order model is used to characterize the corresponding relationship of candidate value score set and probability value set.As an example, The above order model can be the mapping table of the corresponding relationship for characterizing candidate value score set and probability value set. A large amount of candidate value score set and corresponding probability value set are can store in the mapping table.Above-mentioned executing subject It can be similar with the candidate value score set of input (such as between two be calculated gather from being searched in the mapping table Euclidean distance be less than preset distance threshold) candidate value score set, and export the candidate value score collection found Close corresponding probability value set.

In some optional implementations of the present embodiment, above-mentioned executing subject or other electronic equipments can be pressed in advance Order model is obtained according to following steps training:

Firstly, obtaining sample attribute information aggregate.Wherein, sample attribute information corresponds to sample product.Specifically, it is used for The executing subject of training order model can be from long-range or from local obtain sample attribute information aggregate.

Then, for the sample attribute information in sample attribute information aggregate, following steps are executed:

Step 1 is based on the sample attribute information, determines the corresponding sample of sample attribute information using above-mentioned algorithm set This candidate is worth score set.Wherein, determine that the corresponding sample candidate of sample attribute information is worth score collection using algorithm set The method of conjunction can be identical as the method that above-mentioned steps 202 describe, and which is not described herein again.

Step 2 is worth score set and preset, corresponding with the sample attribute information based on obtained sample candidate Value score section, generate at least one markup information.Wherein, markup information corresponds to the algorithm in algorithm set, mark Information is used to characterize the sample candidate determined using corresponding algorithm and is worth whether score is located at the corresponding value point of sample product Number interval.The corresponding of the corresponding relationship and the algorithm in markup information and algorithm set in markup information and value score section is closed System can be pre-set (such as by technical staff's manual setting).Above-mentioned value score section can be the area marked in advance Between.Markup information can be the forms such as number, text, symbol or combinations thereof.For example, markup information is " 1 ", indicate to utilize correspondence Algorithm determine sample candidate be worth score be located at the corresponding value score section of sample product；Markup information is " 0 ", is indicated The sample candidate determined using corresponding algorithm is worth score and is not located at the corresponding value score section of sample product.

It is in section [0,1] as an example it is supposed that the sample candidate of sample product is worth score, the value score of mark Section may include: [0,0.3], and (0.3,0.4], (0.4,0.6], (0.6,0.7], (0.7,1], if sample product corresponds to valence Being worth score section is [0,0.3], and being worth score according to the sample candidate that the corresponding algorithm of certain markup information determines is 0.2, then should Markup information is " 1 ".It should be noted that output is probability value set in the model after actual use is trained, it is each general Rate value corresponds to an algorithm (i.e. the corresponding algorithm of markup information), for characterizing the candidate value score for using the algorithm to determine Probability positioned at the value score section that the corresponding product of candidate value score set for input marks in advance.Probability value can To be the intermediate result exported by order model, rather than the output par, c that order model includes (exports markup information when training Part) output result.

Finally, the corresponding sample of sample attribute information in sample attribute information aggregate is waited using machine learning method Choosing value score set is defeated as it is expected using markup information corresponding with the sample candidate of input value score set as input Out, training obtains order model.

Specifically, the above order model can be the model being trained to initial model.Initial model can wrap Include various existing models, such as neural network model, decision-tree model etc. for classification.Initial model can be set just Beginning parameter, parameter can be continuously adjusted in the training process.For training the executing subject of order model can be based on pre- If loss function calculate penalty values, determine whether initial model trains completion according to penalty values.Herein, it needs to illustrate It is that penalty values can be used for characterizing the difference between reality output and desired output.It (such as is damaged when penalty values meet preset condition Mistake value is less than or equal to preset penalty values threshold value, or after successive ignition calculates, penalty values no longer reduce) when, determine model Training is completed.

Markup information in some optional implementations of the present embodiment, at least one markup information generated Corresponding algorithm determines in accordance with the following steps in advance:

Firstly, for the algorithm in algorithm set, determine that the sample candidate according to determined by the algorithm is worth the standard of score True rate.Specifically, accuracy rate can determine in accordance with the following steps: determine the sample that preset sample product is concentrated according to the algorithm The sample candidate of product is worth score, and accurate sample candidate value score is accounted for determined sample candidate and is worth score The ratio of sum is determined as the accuracy rate that sample candidate determined by the algorithm is worth score.Wherein, accurate sample candidate valence Value score is that the sample candidate in the value score section marked in advance for its corresponding sample product is worth score.

Then, based on identified accuracy rate, selection algorithm is as algorithm corresponding with markup information from algorithm set. Specifically, as an example, for train the executing subject of order model can select accuracy rate highest from algorithm set to A few algorithm, as generating algorithm used by markup information.

As another example, for training the executing subject of order model that can carry out in advance to the algorithm in algorithm set It divides, obtains at least two subclass.For example, algorithm one described in above-mentioned optional embodiment, algorithm can be halved The various calculation methods for not including are determined as a subset conjunction.For each subset in obtained at least two subclass It closes, the highest algorithm of accuracy rate is determined from the subclass.To select at least two algorithms.Can by it is selected at least Two algorithms are used as generating algorithm used by markup information, can also be selected at random from selected at least two algorithm At least one algorithm is selected to be used as generating algorithm used by markup information.

In some optional implementations of the present embodiment, the above order model is trained based on Random Forest model The multi-tag disaggregated model arrived.Multi-tag classification therein refers to that the data of input can be determined that multiple classifications, each Classification corresponds to a kind of label.Each probability in the probability value set of the above order model output, can be used as characterizing The probability of classification belonging to the data of input.In addition, random forest as common classifier, can effectively evade over-fitting and There is noise in data, have very strong robustness.

Step 204, the selection target probability value from probability value set, and will be true according to the corresponding algorithm of destination probability value Fixed candidate value score is determined as label and the storage of target product.

In the present embodiment, above-mentioned executing subject can first from probability value set selection target probability value.As showing Example, above-mentioned executing subject can randomly choose probability value as destination probability value from probability value set；Alternatively, above-mentioned execution master Body can select most probable value as destination probability value from probability value set.

Then, above-mentioned executing subject can determine the candidate value score determined according to the corresponding algorithm of destination probability value Label and storage for target product.Specifically, identified label can store in the memory block of above-mentioned executing subject local, It is stored in the memory block with other electronic equipments of above-mentioned executing subject communication connection.Wherein, it should be noted that on Stating memory block can be memory block of example, in hardware, such as hard disk, memory etc..It is also possible to the memory block of software form, such as Database, list etc..

Optionally, the label of target product can also be further illustrated on the display connecting with above-mentioned executing subject.

By executing above steps, due to using order model, identified label can be made accurately to characterize The value of target product.Facilitate the information for targetedly pushing characterization target product to user terminal according to label.

With continued reference to the signal that Fig. 3, Fig. 3 are according to the application scenarios of the method for generating information of the present embodiment Figure.In the application scenarios of Fig. 3, server 301 obtains the attribute information 302 of target product (such as certain mobile phone) first.Its In, attribute information includes the property of value value 3021 (such as price " 2000 ") of target product.Then, server 301 is using in advance If algorithm set 303 (for example including algorithm a1, a2 ..., a22, each algorithm characterizes in the form of application program) in it is every A algorithm calculates property of value value 3021, obtains candidate value score set 304.Subsequently, candidate is worth score Set 304 with vector (such as the vector of 22 dimensions, including numerical value n1, n2 ..., n22) the trained in advance order mould of form input Type 305 obtains probability value set 306 (for example including 6 probability values 0.65,0.88,0.84,0.41,0.54,0.63).Wherein, Each probability value corresponds to an algorithm in algorithm set, for characterizing the candidate value score determined using corresponding algorithm Order of accuarcy.Then, server 301 selects maximum value (i.e. 0.88) as destination probability value, target from probability value set Probability value corresponds to the algorithm a10 in algorithm set.Finally using the candidate value score n10 determined according to algorithm a10 as mesh The label of product is marked, and in the memory block 307 (such as hard disk) for including to server 301 by label storage.

The method provided by the above embodiment of the application passes through attribute information based on target product and preset set of algorithms It closes, determines candidate value score set, then candidate value score set input order model trained in advance is obtained into probability value Set, finally the selection target probability value from probability value set, and the time that will be determined according to the corresponding algorithm of destination probability value Choosing value score is determined as label and the storage of target product, to be improved using candidate value score set and order model The accuracy for generating the label of target product facilitates the specific aim that recommendation information is improved using the label generated.

With further reference to Fig. 4, it illustrates the processes 400 of another embodiment of the method for generating information.The use In the process 400 for the method for generating information, comprising the following steps:

Step 401, the attribute information of target product is obtained.

In the present embodiment, can lead to for generating the executing subject (such as server shown in FIG. 1) of the method for information Wired connection mode or radio connection are crossed from attribute information long-range or from local target product.Wherein, target product It can be the product of its label to be determined, such as the product of the product information characterization of certain website offer be provided.Attribute information can To include information relevant to each attribute of target product, such as title, weight, size, the characterization target of target product produce Property of value value of the property of value of product etc..

Step 402, determine whether the order volume of target product in the target time period is more than or equal to preset order volume threshold Value.

In the present embodiment, the attribute information of target product includes the order volume of target product in the target time period.On Stating executing subject can determine whether the order volume of target product in the target time period is more than or equal to preset order volume threshold value (such as 10).

Wherein, target time section is that above-mentioned executing subject is determined according to the pre-set time method of determination of technical staff Period.For example, target time section can be nearest one day, one week, the periods such as January.Order volume is mesh in target time section The number that mark product is acquired.

In the present embodiment, above-mentioned executing subject can be in response to determining the order volume of target product in the target time period More than or equal to the above order amount threshold value, step 403- step 404 is executed, in response to determining that being more than or equal to the above order amount threshold value holds Row step 405- step 406.

Step 403, it is more than or equal to order volume threshold value in response to determining, based in attribute information and preset algorithm set Algorithm determines candidate value score set as the first candidate value score set.

In the present embodiment, above-mentioned executing subject can be in response to determining the order volume of target product in the target time period Candidate value score set is determined based on the algorithm in attribute information and preset algorithm set more than or equal to order volume threshold value As the first candidate value score set.

Specifically, (such as Fig. 2 is corresponding for the numerical value for the property of value for characterizing target product that attribute information may include Real value attribute value and original value attribute value described in embodiment), i.e., the numerical value obtains target production for characterizing user The cost that product are paid.For each algorithm in above-mentioned algorithm set, include using the algorithm and above-mentioned attribute information For characterizing the numerical value of the property of value of target product, the corresponding candidate value score of the algorithm can be calculated.Thus will Obtained candidate value score set is determined as the first candidate value score set.

It step 404, will in response to determining that the order volume of target product in the target time period is more than or equal to order volume threshold value The first model that first candidate value score set input order model includes, obtains probability value set.

In the present embodiment, above-mentioned executing subject can be in response to determining the order volume of target product in the target time period More than or equal to order volume threshold value, the first model for including by the first candidate value score set input order model obtains probability Value set.

Specifically, above-mentioned first model is used to characterize the corresponding relationship of candidate value score set and probability value set.Make For example, above-mentioned first model can be corresponding with the corresponding relationship of probability value set for characterizing candidate value score set Relation table.Alternatively, above-mentioned first model can be using preset training sample set, based on the existing mould for classification Type, the model obtained using machine learning method, training.It should be noted that training the first model method can with it is above-mentioned The method of training order model described in optional implementation in Fig. 2 embodiment is identical, and which is not described herein again.

Step 405, in response to determine be less than order volume threshold value, based on it is in attribute information and preset algorithm set, remove Algorithm other than target algorithm determines candidate value score set as the second candidate value score set.

In the present embodiment, above-mentioned executing subject can be in response to determining the order volume of target product in the target time period It is determined less than order volume threshold value based on other algorithms in attribute information and preset algorithm set, in addition to target algorithm Candidate's value score set is as the second candidate value score set.Wherein, target algorithm is and ordering in the target time period Single amount is more than or equal to the corresponding algorithm of product of order volume threshold value, i.e., target algorithm is for determining order in the target time period Amount is more than or equal to the candidate value score of the product of order volume threshold value.

As an example it is supposed that algorithm set includes algorithm A, B, C, D, E, wherein A, B are arranged to target algorithm.If mesh It marks the order volume of product in the target time period and is more than or equal to order volume threshold value, then determined using each algorithm in algorithm set Candidate's value score set is as the first candidate value score set；Otherwise, candidate value score collection is determined using algorithm C, D, E Cooperation is the second candidate value score set.

It is two parts by the algorithm partition in algorithm set by the size according to order volume, can determining time When choosing value score set, various forms of candidate value score collection are targetedly obtained according to the size of the order volume of product It closes, to help to improve the accuracy of the label of product.

In some optional implementations of the present embodiment, target algorithm may include:

Firstly, in response to determining that the order volume of target product in the target time period is more than or equal to preset order volume threshold Value obtains product that is predetermined, obtaining the user in the user group of target product and obtains correlation.Wherein, product obtains It takes correlation and can be predetermined, for characterizing cost that user pays in the target time period in order to obtain product How many numerical value, product obtain the cost that correlation is bigger, and expression user pays in the target time period in order to obtain product It is bigger, and then the value that can characterize product indirectly is bigger.It can be for example, product obtains correlation in [0,1] section In numerical value.It, can also be with it should be noted that the product of user, which obtains correlation, can be the numerical value of technical staff's manual setting It is the numerical value that above-mentioned executing subject or other electronic equipments are determined according to various methods.As an example, can be according to existing number According to method for normalizing (such as standard normalization, minimax normalization etc.), by the user in above-mentioned user group in the object time The cost value that product is paid is obtained in section, is mapped to [0,1] section, so that the product for obtaining each user obtains correlation. It may be embodied in user's portrait information of predetermined feature for characterizing user in general, product obtains correlation.

Then, correlation is obtained based on obtained product, determines the candidate value score of target product.As an example, Above-mentioned executing subject can obtain in correlation from the product of the user in above-mentioned user group, select median as target product Candidate value score.Alternatively, the mean value that each product obtains correlation can be determined as target product by above-mentioned executing subject Candidate value score.

In the case that order volume in the target time period is more than or equal to order volume threshold value, using target algorithm, Ke Yiyou Effect obtains correlation using the product of user, realizes the value score set that information related with user is dissolved into product, from And be conducive to improve the accuracy for the label for generating product.

In some optional implementations of the present embodiment, target algorithm may include:

It is first in response to determining that the order volume of target product in the target time period is more than or equal to preset order volume threshold value First, from preset at least two value score section, at least one following value score section: the reality of target product is determined The property of value is worth the corresponding value score section of original value attribute value in corresponding value score section, target product.Wherein, Preset at least two value score section is the value score section that technical staff marks in advance, real value attribute value and valence It is worth the corresponding relationship in score section, it is pre- that original value attribute value and the corresponding relationship for being worth score section are also possible to technical staff First it is arranged.As an example it is supposed that the candidate value score of product is in section [0,1], preset at least two value point Number interval may include: [0,0.4], and (0.4,0.5], (0.5,0.8], (0.8,0.9], (0.9,1].On it should be noted that Stating at least two value score sections may include two groups of value score sections, correspond respectively to real value attribute value and original Property of value value.Above-mentioned at least two value score section also may include one group of value score section, real value attribute value Group value score section is both corresponded to original value attribute value.

Then, it for the value score section at least one identified value score section, obtains preset and is somebody's turn to do It is worth the corresponding transformation coefficient in score section, and determines target candidate from the candidate value score for the target product determined It is worth score.As an example, above-mentioned executing subject can be according to the specified of technical staff, from the candidate of fixed target product It is worth in score and determines that target candidate is worth score.For example, can be by the example in above-mentioned algorithm one and algorithm two, basis Candidate determined by the product information sequence that the corresponding product information of each order is ranked up is worth score as mesh The candidate value score of mark.Above-mentioned executing subject then can be worth score based on identified transformation coefficient and target candidate, really Set the goal the new candidate value score of product.As an example, above-mentioned executing subject can be by identified transformation coefficient and mesh The product that the candidate value fractional multiplication of mark obtains is determined as new candidate value score.Alternatively, can be on the basis of above-mentioned product On converted (such as be added or subtract each other with setting value), obtain new candidate value score.In general, can will be obtained new Candidate value score replace former candidate value score.

It, can order volume in the target time period the case where being more than or equal to order volume threshold value by executing this implementation Under, obtained candidate value score is transformed into new value score section from locating preset value score section, So as to neatly adjust value interval locating for candidate value score convenient for technical staff.

Step 406, in response to determining that the order volume of target product in the target time period is less than order volume threshold value, by second The second model that candidate's value score set input order model includes, obtains probability value set.

In the present embodiment, above-mentioned executing subject can be in response to determining the order volume of target product in the target time period Less than order volume threshold value, the second model for including by the second candidate value score set input order model obtains probability value collection It closes.

Specifically, above-mentioned second model is used to characterize the corresponding relationship of candidate value score set and probability value set.Make For example, above-mentioned second model can be corresponding with the corresponding relationship of probability value set for characterizing candidate value score set Relation table.Alternatively, above-mentioned second model can be using preset training sample set, based on the existing mould for classification Type, the model obtained using machine learning method, training.

It should be noted that the method for the second model of training can be similar with the training method of the first model, unlike, It does not include target described in above-mentioned optional implementation that sample candidate used in the second model of training, which is worth score set, Algorithm, in turn, generating algorithm used in markup information used in the second model of training does not also include target algorithm.About instruction Practice the method for the second model, the training order model with reference to described in the optional implementation in above-mentioned Fig. 2 embodiment Method, which is not described herein again.

Step 407, the selection target probability value from probability value set, and will be true according to the corresponding algorithm of destination probability value Fixed candidate value score is determined as label and the storage of target product.

In the present embodiment, step 407 and the step 204 in Fig. 2 corresponding embodiment are almost the same, and which is not described herein again.

Figure 4, it is seen that the method for generating information compared with the corresponding embodiment of Fig. 2, in the present embodiment Process 400 highlight whether the order volume of determining target product in the target time period is more than or equal to preset order volume threshold Value, and the step of using the first model or the second model generating probability value set.The scheme of the present embodiment description can be with as a result, According to the size of the order volume of target product in the target time period, different processing modes is taken, obtains the mark of target product Label, to further improve the accuracy for generating the label of target product.

With further reference to Fig. 5, it illustrates the processes 500 of one embodiment of the method for pushed information.This is used for The process 500 of the method for pushed information, comprising the following steps:

Step 501, the classification information and preset product information set of target user are obtained.

In the present embodiment, can lead to for the executing subject of the method for pushed information (such as server shown in FIG. 1) Wired connection mode or radio connection are crossed from classification information and preset production long-range or from local acquisition target user Product information aggregate.Wherein, the product information in product information set corresponds to predetermined label.Label is according to above-mentioned figure What method described in 2 or Fig. 4 corresponding embodiment generated.Target user can be push product information of the terminal to use to it User.The classification information of target user can be used for characterizing the classification of user group locating for target user, and classification information can be with Information including following at least one form: number, text, symbol etc..

As an example, label can be the numerical value in [0,1] section, which is divided at least two sub-districts in advance Between, for example, [0,0.3], (0.3,0.4], (0.4,0.6], (0.6,0.7], (0.7,1].Each subinterval can correspond to In a classification information.It should be noted that classification information can be previously obtained in various manners, for example, can have technology Personnel's manual setting.Alternatively, the cost value that can be paid in the target time period each user in preset user group It is counted, determines the section of cost value, which is divided, obtain at least two subintervals, each subinterval is corresponding In an above-mentioned subinterval for label, the corresponding classification information in each subinterval is generated, to can determine that user is corresponding Classification information.

Product information can be the information for characterizing product, may include following at least one information: the number of product, Image, the place of production, property of value value etc..It should be appreciated that label corresponding with product information may include in product information, it can also The corresponding relationship of product information and label is established in the form of through two-dimensional table, chained list etc..

Step 502, the corresponding relationship based on preset, label and classification information, from product information set, determining and class The product information of other information matches.

In the present embodiment, above-mentioned executing subject can the corresponding relationship based on preset, label and classification information, from production In product information aggregate, the determining and matched product information of classification information.

Specifically, as an example, label and the corresponding relationship of classification information can pass through the forms table such as two-dimensional table, chained list Sign.Above-mentioned executing subject can determine that corresponding label is corresponding with the classification information of target user from product information set Product information as with the matched product information of classification information.

As another example, above-mentioned executing subject can determine corresponding label and mesh first from product information set The corresponding product information of classification information for marking user, then from identified product information, selection target product information as with The matched product information of classification information.For example, product information may include order volume, when above-mentioned executing subject can choose target Between in section the maximum preset quantity product information of order volume as target product information.Alternatively, above-mentioned executing subject can root Classification belonging to the product obtained according to target user, from identified product information, selection belongs to such purpose product Information is as target product information.

Step 503, identified product information is pushed to the terminal of target user.

In the present embodiment, product information identified in step 502 can be pushed to target use by above-mentioned executing subject The terminal at family, so that target user browses product information by the terminal that it is used.

The method provided by the above embodiment of the application is believed by the classification information and preset product that obtain target user Breath set, is then based on the corresponding relationship of preset, label and classification information, determining to believe with classification from product information set Matched product information is ceased, identified product information is finally pushed to the terminal of target user, by embodiment institute benefit Label can accurately characterize the value of target product, therefore being directed to the terminal pushed information of user can be improved Property.

With further reference to Fig. 6, as the realization to method shown in above-mentioned Fig. 2, this application provides one kind for generating letter One embodiment of the device of breath, the Installation practice is corresponding with embodiment of the method shown in Fig. 2, which can specifically answer For in various electronic equipments.

As shown in fig. 6, the present embodiment includes: acquiring unit 601 for generating the device 600 of information, it is configured to obtain Take the attribute information of target product；Determination unit 602 is configured to be determined and waited based on attribute information and preset algorithm set Choosing value score set；Generation unit 603 is configured to the order model that candidate value score set input is trained in advance, Obtain probability value set, wherein the probability value in probability value set corresponds to the algorithm in algorithm set, for characterizing utilization pair The order of accuarcy for the candidate value score that the algorithm answered determines；Storage unit 604 is configured to select mesh from probability value set Probability value is marked, and the candidate value score determined according to the corresponding algorithm of destination probability value is determined as to the label of target product And storage.

In the present embodiment, acquiring unit 601 can by wired connection mode or radio connection from long-range or From the local attribute information for obtaining target product.Wherein, target product can be the product of its label to be determined, such as can be The product for the product information characterization that certain website provides.Attribute information may include letter relevant to each attribute of target product Breath.For example, attribute information may include following at least one information: title, weight, size, the characterization target of target product produce Property of value value of the property of value of product etc..

In the present embodiment, determination unit 602 can determine candidate value score set.Wherein, candidate value score can With the size of the value for characterizing target product.

Specifically, attribute information includes the numerical value for characterizing the property of value of target product, which can be used for table Requisition family obtains the cost that target product is paid.For the algorithm in above-mentioned algorithm set, the algorithm and above-mentioned category are utilized Property the information property of value for characterizing target product that includes numerical value, the corresponding candidate value of the algorithm can be calculated Score.In general, the algorithm in algorithm set can be characterized in the form of formula or application program, the value of target product will be characterized The numerical value of attribute is input to formula or application program, available candidate value score.

In the present embodiment, the order model that generation unit 603 can be trained in advance by candidate value score set input, Obtain probability value set.Wherein, the probability value in probability value set corresponds to the algorithm in algorithm set, for characterizing utilization pair The order of accuarcy for the candidate value score that the algorithm answered determines.

In general, candidate value score set can input order model, the probability value set of output in vector form It can be the form of vector.The quantity for the probability value that probability value set includes is less than the candidate valence that candidate value score set includes It is worth the quantity of score.Each probability value corresponds to an algorithm in above-mentioned algorithm set.Probability value is higher, indicates general according to this The accuracy that rate is worth the candidate value score that corresponding algorithm determines is higher.

The above order model is used to characterize the corresponding relationship of candidate value score set and probability value set.As an example, The above order model can be the mapping table of the corresponding relationship for characterizing candidate value score set and probability value set. A large amount of candidate value score set and corresponding probability value set are can store in the mapping table.Above-mentioned generation unit 603 can search (such as two set being calculated similar to the candidate value score set of input from the mapping table Between Euclidean distance be less than preset distance threshold) candidate value score set, and export the candidate value point found Manifold closes corresponding probability value set.

In the present embodiment, storage unit 604 can first from probability value set selection target probability value.As showing Example, said memory cells 604 can randomly choose probability value as destination probability value from probability value set.In general, above-mentioned deposit Storage unit 604 can select maximum value as destination probability value from probability value set.

Then, said memory cells 604 can first from probability value set selection target probability value.On as an example, Probability value can be randomly choosed as destination probability value from probability value set by stating storage unit 604；Alternatively, above-mentioned executing subject Most probable value can be selected as destination probability value from probability value set.Then, storage unit 604 can will be according to target The candidate value score that the corresponding algorithm of probability value determines is determined as label and the storage of target product.Specifically, identified Label can store the memory block local in above-mentioned apparatus 600, be stored in other communicated to connect with above-mentioned apparatus 600 The memory block of electronic equipment.Wherein, it should be noted that above-mentioned memory block can be the memory block of example, in hardware, such as hard disk, Memory etc..It is also possible to memory block of software form, such as database, list etc..

In some optional implementations of the present embodiment, attribute information may include target product in target time section Interior order volume；And determination unit 602 may include: the first determining module (not shown), be configured to determine target Whether the order volume of product in the target time period is more than or equal to preset order volume threshold value；Second determining module (is not shown in figure Out), it is configured in response to determine and is more than or equal to order volume threshold value, based on the calculation in attribute information and preset algorithm set Method determines candidate value score set as the first candidate value score set；Third determining module (not shown), is matched Be set in response to determine be less than order volume threshold value, based on it is in attribute information and preset algorithm set, in addition to target algorithm Other algorithms, determine candidate value score set as the second candidate value score set, wherein target algorithm be in mesh Mark the corresponding algorithm of product that the order volume in the period is more than or equal to order volume threshold value.

In some optional implementations of the present embodiment, target algorithm includes: in response to determining target product in mesh The order volume marked in the period is more than or equal to preset order volume threshold value, obtains use that is predetermined, obtaining target product The product of user in the group of family obtains correlation；Correlation is obtained based on obtained product, determines the candidate valence of target product It is worth score.

In some optional implementations of the present embodiment, target algorithm includes: in response to determining target product in mesh The order volume marked in the period is more than or equal to preset order volume threshold value, from preset at least two value score section, really At least one fixed following value score section: the corresponding value score section of the real value attribute value of target product, target produce The corresponding value score section of the original value attribute value of product；For the value at least one identified value score section Score section obtains transformation coefficient preset, corresponding with the value score section, and the time from fixed target product Determine that target candidate is worth score in choosing value score；It is worth score based on identified transformation coefficient and target candidate, is determined The new candidate value score of target product.

In some optional implementations of the present embodiment, order model includes the first model and the second model；And Generation unit 603 may include: the first generation module (not shown), be configured in response to determine target product in target Order volume in period is more than or equal to order volume threshold value, and the first candidate value score set is inputted the first model, obtain to Few two probability values；Second generation module (not shown) is configured in response to determine target product in target time section Interior order volume is less than order volume threshold value, and the second candidate value score set is inputted the second model, obtains at least two probability Value.

In some optional implementations of the present embodiment, attribute information comprises at least one of the following numerical value: target produces The original value attribute value of the real value attribute values of product, target product；And algorithm set comprises at least one of the following algorithm: Real value attribute value based on the product that classification belonging to target product includes, the production for including to classification belonging to target product The product information of product is ranked up, and obtains at least two first product information sequences, is produced for obtained at least two first The first product information sequence in product information sequence, the arrangement position based on target product in the first product information sequence, Determine that target product corresponds to the candidate value score of the first product information sequence；Include based on classification belonging to target product Product original value attribute value, the product information for the product that classification belonging to target product includes is ranked up, is obtained At least two second product information sequences, for the second product information in obtained at least two second product information sequence Sequence, the arrangement position based on target product in the second product information sequence determine that target product corresponds to second production The candidate value score of product information sequence.

In some optional implementations of the present embodiment, algorithm set can also comprise at least one of the following algorithm:

One, the mean value and standard deviation of the real value attribute value for the product that classification belonging to target product includes are determined；Base In identified mean value and standard deviation, the candidate value score of target product is determined.

Two, the mean value and standard deviation of the original value attribute value for the product that classification belonging to target product includes are determined；Base In identified mean value and standard deviation, the candidate value score of target product is determined.

In some optional implementations of the present embodiment, order model can be trained in accordance with the following steps in advance It arrives: obtaining sample attribute information aggregate, wherein sample attribute information corresponds to sample product；For sample attribute information aggregate In sample attribute information, be based on the sample attribute information, determine the corresponding sample of sample attribute information using algorithm set Candidate's value score set；Score set and the preset and sample attribute information pair are worth based on obtained sample candidate The value score section answered, generates at least one markup information, wherein markup information corresponds to the algorithm in algorithm set, mark Note information is used to characterize the sample candidate determined using corresponding algorithm and is worth whether score is located at the corresponding value of sample product Score section；Using machine learning method, by the corresponding sample candidate valence of sample attribute information in sample attribute information aggregate It is worth score set as input, is worth the corresponding markup information of score set as desired output for the sample candidate of input, Training obtains order model.

Markup information in some optional implementations of the present embodiment, at least one markup information generated Corresponding algorithm can determine in accordance with the following steps in advance: for the algorithm in algorithm set, determine according to the algorithm institute Determining sample candidate is worth the accuracy rate of score；Based on identified accuracy rate, from algorithm set selection algorithm as with The corresponding algorithm of markup information.

In some optional implementations of the present embodiment, order model is obtained based on Random Forest model training Multi-tag disaggregated model.

The device provided by the above embodiment of the application passes through attribute information based on target product and preset set of algorithms It closes, determines candidate value score set, then candidate value score set input order model trained in advance is obtained into probability value Set, finally the selection target probability value from probability value set, and the time that will be determined according to the corresponding algorithm of destination probability value Choosing value score is determined as label and the storage of target product, to be improved using candidate value score set and order model The accuracy for generating the label of target product facilitates the specific aim that recommendation information is improved using the label generated.

With further reference to Fig. 7, as the realization to method shown in above-mentioned Fig. 5, this application provides one kind for pushing letter One embodiment of the device of breath, the Installation practice is corresponding with embodiment of the method shown in fig. 5, which can specifically answer For in various electronic equipments.

As shown in fig. 7, the device 700 for pushed information of the present embodiment includes: acquiring unit 701, it is configured to obtain Take the classification information and preset product information set of target user, wherein product information corresponds to predetermined label, mark Label are that the method according to described in above-mentioned Fig. 2 or Fig. 4 corresponding embodiment generates；Determination unit 702 is configured to based on default , the corresponding relationship of label and classification information, it is determining with the matched product information of classification information from product information set；It pushes away Unit 703 is sent, is configured to push to identified product information the terminal of target user.

In the present embodiment, acquiring unit 701 can by wired connection mode or radio connection from long-range or From the local classification information and preset product information set for obtaining target user.Wherein, the product letter in product information set Breath corresponds to predetermined label.Label is that the method according to described in above-mentioned Fig. 2 or Fig. 4 corresponding embodiment generates.Mesh Mark user can be the user of push product information of the terminal to use to it.The classification information of target user can be used for characterizing The classification of user group locating for target user, classification information may include the information of following at least one form: number, text, Symbol etc..As an example, label can be the numerical value in [0,1] section, which is divided at least two sub-districts in advance Between, for example, [0,0.3], (0.3,0.4], (0.4,0.6], (0.6,0.7], (0.7,1].Each subinterval can correspond to In a classification information.It should be noted that classification information can be previously obtained in various manners, for example, can have technology Personnel's manual setting.Alternatively, the cost value that can be paid in the target time period each user in preset user group It is counted, determines the section of cost value, which is divided, obtain at least two subintervals, each subinterval is corresponding In an above-mentioned subinterval for label, the corresponding classification information in each subinterval is generated, to can determine that user is corresponding Classification information.

Product information can be the information for characterizing product, may include following at least one information: the number of product, Image, the place of production, property of value value etc..It should be appreciated that label corresponding with product information may include in product information, it can also The corresponding relationship of product information and label is established in the form of through two-dimensional table, chained list etc..

In the present embodiment, determination unit 702 can the corresponding relationship based on preset, label and classification information, from production In product information aggregate, the determining and matched product information of classification information.

Specifically, as an example, label and the corresponding relationship of classification information can pass through the forms table such as two-dimensional table, chained list Sign.Above-mentioned executing subject can determine that corresponding label is corresponding with the classification information of target user from product information set Product information as with the matched product information of classification information.

As another example, above-mentioned executing subject can determine corresponding label and mesh first from product information set The corresponding product information of classification information for marking user, then from identified product information, selection target product information as with The matched product information of classification information.For example, product information may include order volume, when above-mentioned executing subject can choose target Between in section the maximum preset quantity product information of order volume as target product information.Alternatively, above-mentioned executing subject can root Classification belonging to the product obtained according to target user, from identified product information, selection belongs to such purpose product Information is as target product information.

In the present embodiment, product information determined by determination unit 702 can be pushed to target use by push unit 703 The terminal at family, so that target user browses product information by the terminal that it is used.

The device provided by the above embodiment of the application is believed by the classification information and preset product that obtain target user Breath set, is then based on the corresponding relationship of preset, label and classification information, determining to believe with classification from product information set Matched product information is ceased, identified product information is finally pushed to the terminal of target user, by embodiment institute benefit Label can accurately characterize the value of target product, therefore being directed to the terminal pushed information of user can be improved Property.

Below with reference to Fig. 8, it illustrates the computer systems 800 for the server for being suitable for being used to realize the embodiment of the present application Structural schematic diagram.Server shown in Fig. 8 is only an example, should not function and use scope band to the embodiment of the present application Carry out any restrictions.

As shown in figure 8, computer system 800 includes central processing unit (CPU) 801, it can be read-only according to being stored in Program in memory (ROM) 802 or be loaded into the program in random access storage device (RAM) 803 from storage section 808 and Execute various movements appropriate and processing.In RAM 803, also it is stored with system 800 and operates required various programs and data. CPU 801, ROM 802 and RAM 803 are connected with each other by bus 804.Input/output (I/O) interface 805 is also connected to always Line 804.

I/O interface 805 is connected to lower component: the importation 806 including keyboard, mouse etc.；Including such as liquid crystal Show the output par, c 807 of device (LCD) etc. and loudspeaker etc.；Storage section 808 including hard disk etc.；And including such as LAN The communications portion 809 of the network interface card of card, modem etc..Communications portion 809 is executed via the network of such as internet Communication process.Driver 810, which also can according to need, is connected to I/O interface 805.Detachable media 811, such as disk, CD, Magneto-optic disk, semiconductor memory etc. are mounted on as needed on driver 810, in order to from the computer journey read thereon Sequence is mounted into storage section 808 as needed.

Particularly, in accordance with an embodiment of the present disclosure, it may be implemented as computer above with reference to the process of flow chart description Software program.For example, embodiment of the disclosure includes a kind of computer program product comprising be carried on computer-readable medium On computer program, which includes the program code for method shown in execution flow chart.In such reality It applies in example, which can be downloaded and installed from network by communications portion 809, and/or from detachable media 811 are mounted.When the computer program is executed by central processing unit (CPU) 801, limited in execution the present processes Above-mentioned function.

It should be noted that computer-readable medium described herein can be computer-readable signal media or meter Calculation machine readable medium either the two any combination.Computer-readable medium for example may be-but not limited to- Electricity, magnetic, optical, electromagnetic, infrared ray or semiconductor system, device or device, or any above combination.It is computer-readable The more specific example of medium can include but is not limited to: have electrical connection, the portable computer magnetic of one or more conducting wires Disk, hard disk, random access storage device (RAM), read-only memory (ROM), erasable programmable read only memory (EPROM or sudden strain of a muscle Deposit), optical fiber, portable compact disc read-only memory (CD-ROM), light storage device, magnetic memory device or above-mentioned appoint The suitable combination of meaning.In this application, computer-readable medium can be any tangible medium for including or store program, the journey Sequence can be commanded execution system, device or device use or in connection.And in this application, it is computer-readable Signal media may include in a base band or as carrier wave a part propagate data-signal, wherein carrying computer can The program code of reading.The data-signal of this propagation can take various forms, including but not limited to electromagnetic signal, optical signal or Above-mentioned any appropriate combination.Computer-readable signal media can also be any calculating other than computer-readable medium Machine readable medium, the computer-readable medium can be sent, propagated or transmitted for by instruction execution system, device or device Part uses or program in connection.The program code for including on computer-readable medium can use any Jie appropriate Matter transmission, including but not limited to: wireless, electric wire, optical cable, RF etc. or above-mentioned any appropriate combination.

The calculating of the operation for executing the application can be write with one or more programming languages or combinations thereof Machine program code, described program design language include object oriented program language-such as Java, Smalltalk, C+ +, it further include conventional procedural programming language-such as " C " language or similar programming language.Program code can Fully to execute, partly execute on the user computer on the user computer, be executed as an independent software package, Part executes on the remote computer or executes on a remote computer or server completely on the user computer for part. In situations involving remote computers, remote computer can pass through the network of any kind --- including local area network (LAN) Or wide area network (WAN)-is connected to subscriber computer, or, it may be connected to outer computer (such as utilize Internet service Provider is connected by internet).

Flow chart and block diagram in attached drawing are illustrated according to the system of the various embodiments of the application, method and computer journey The architecture, function and operation in the cards of sequence product.In this regard, each box in flowchart or block diagram can generation A part of one module, program segment or code of table, a part of the module, program segment or code include one or more use The executable instruction of the logic function as defined in realizing.It should also be noted that in some implementations as replacements, being marked in box The function of note can also occur in a different order than that indicated in the drawings.For example, two boxes succeedingly indicated are actually It can be basically executed in parallel, they can also be executed in the opposite order sometimes, and this depends on the function involved.Also it to infuse Meaning, the combination of each box in block diagram and or flow chart and the box in block diagram and or flow chart can be with holding The dedicated hardware based system of functions or operations as defined in row is realized, or can use specialized hardware and computer instruction Combination realize.

Being described in unit involved in the embodiment of the present application can be realized by way of software, can also be by hard The mode of part is realized.Described unit also can be set in the processor, for example, can be described as: a kind of processor, packet Include acquiring unit, determination unit, generation unit and storage unit.Wherein, the title of these units not structure under certain conditions The restriction of the pairs of unit itself, for example, acquiring unit is also described as " obtaining the list of the attribute information of target product Member ".

As on the other hand, present invention also provides a kind of computer-readable medium, which be can be Included in server described in above-described embodiment；It is also possible to individualism, and without in the supplying server.It is above-mentioned Computer-readable medium carries one or more program, when said one or multiple programs are executed by the server, So that the server: obtaining the attribute information of target product；Based on attribute information and preset algorithm set, candidate value is determined Score set；By candidate value score set input order model trained in advance, probability value set is obtained, wherein probability value Probability value in set corresponds to the algorithm in algorithm set, for characterizing the candidate value score determined using corresponding algorithm Order of accuarcy；The selection target probability value from probability value set, and will be according to the corresponding algorithm determination of destination probability value Candidate's value score is determined as label and the storage of target product.

In addition, when said one or multiple programs are executed by the server, it is also possible that the server: obtaining mesh Mark the classification information and preset product information set of user, wherein product information corresponds to predetermined label, and label is It is generated according to the method that any embodiment in above-mentioned first aspect describes；Pair based on preset, label and classification information It should be related to, from product information set, the determining and matched product information of classification information；Identified product information is pushed to The terminal of target user.

Above description is only the preferred embodiment of the application and the explanation to institute's application technology principle.Those skilled in the art Member is it should be appreciated that invention scope involved in the application, however it is not limited to technology made of the specific combination of above-mentioned technical characteristic Scheme, while should also cover in the case where not departing from foregoing invention design, it is carried out by above-mentioned technical characteristic or its equivalent feature Any combination and the other technical solutions formed.Such as features described above has similar function with (but being not limited to) disclosed herein Can technical characteristic replaced mutually and the technical solution that is formed.

Claims (15)

1. a kind of method for generating information, comprising:

Obtain the attribute information of target product；

Based on the attribute information and preset algorithm set, candidate value score set is determined；

By candidate value score set input order model trained in advance, probability value set is obtained, wherein the probability Probability value in value set corresponds to the algorithm in the algorithm set, for characterizing the candidate valence determined using corresponding algorithm It is worth the order of accuarcy of score；

The selection target probability value from the probability value set, and will be according to the corresponding algorithm determination of the destination probability value Candidate's value score is determined as label and the storage of the target product.

2. according to the method described in claim 1, wherein, the attribute information include the target product in the target time period Order volume；And

It is described to be based on the attribute information and preset algorithm set, determine candidate value score set, comprising:

Determine whether the order volume of the target product in the target time period is more than or equal to preset order volume threshold value；

It is more than or equal to the order volume threshold value in response to determining, based on the calculation in the attribute information and preset algorithm set Method determines candidate value score set as the first candidate value score set；

In response to determine be less than the order volume threshold value, based on it is in the attribute information and preset algorithm set, remove target Other algorithms other than algorithm determine candidate value score set as the second candidate value score set, wherein the target Algorithm is algorithm corresponding more than or equal to the product of the order volume threshold value with the order volume in the target time section.

3. according to the method described in claim 2, wherein, the target algorithm includes:

It is more than or equal to preset order volume threshold value in response to the order volume of the determination target product in the target time period, obtains The product of user in user group that is predetermined, obtaining the target product obtains correlation；

Correlation is obtained based on obtained product, determines the candidate value score of the target product.

4. according to the method described in claim 2, wherein, the target algorithm includes:

It is more than or equal to preset order volume threshold value in response to the order volume of the determination target product in the target time period, from pre- If at least two value score sections in, determine at least one following value score section: the real price of the target product The corresponding corresponding value score section of original value attribute value for being worth score section, the target product of value attribute value；

For the value score section at least one identified value score section, the preset and value score is obtained The corresponding transformation coefficient in section, and target candidate value is determined from the candidate value score of the fixed target product Score；It is worth score based on identified transformation coefficient and target candidate, determines the new candidate value point of the target product Number.

5. according to the method described in claim 2, wherein, the order model includes the first model and the second model；And

The order model that the candidate value score set input is trained in advance, obtains probability value set, comprising:

It is more than or equal to the order volume threshold value in response to the order volume of the determination target product in the target time period, it will be described First candidate value score set inputs first model, obtains at least two probability values；

It is less than the order volume threshold value in response to the order volume of the determination target product in the target time period, by described second Candidate's value score set inputs second model, obtains at least two probability values.

6. according to the method described in claim 1, wherein, the attribute information comprises at least one of the following numerical value: the target The original value attribute value of the real value attribute value of product, the target product；And

The algorithm set comprises at least one of the following algorithm:

Real value attribute value based on the product that classification belonging to the target product includes, to belonging to the target product The product information for the product that classification includes is ranked up, and obtains at least two first product information sequences, for it is obtained extremely The first product information sequence in few two the first product information sequences, based on the target product in the first product information sequence Arrangement position in column determines that the target product corresponds to the candidate value score of the first product information sequence；

Original value attribute value based on the product that classification belonging to the target product includes, to belonging to the target product The product information for the product that classification includes is ranked up, and obtains at least two second product information sequences, for it is obtained extremely The second product information sequence in few two the second product information sequences, based on the target product in the second product information sequence Arrangement position in column determines that the target product corresponds to the candidate value score of the second product information sequence.

7. according to the method described in claim 6, wherein, the algorithm set further includes following at least one algorithm:

Determine the mean value and standard deviation of the real value attribute value for the product that classification belonging to the target product includes；Based on institute Determining mean value and standard deviation determines the candidate value score of the target product；

Determine the mean value and standard deviation of the original value attribute value for the product that classification belonging to the target product includes；Based on institute Determining mean value and standard deviation determines the candidate value score of the target product.

8. method described in one of -7 according to claim 1, wherein the order model is trained in accordance with the following steps in advance It arrives:

Obtain sample attribute information aggregate, wherein sample attribute information corresponds to sample product；

For the sample attribute information in the sample attribute information aggregate, it is based on the sample attribute information, utilizes the algorithm Gather and determines that the corresponding sample candidate of the sample attribute information is worth score set；Score is worth based on obtained sample candidate Set and value score section preset, corresponding with the sample attribute information, generate at least one markup information, wherein mark The algorithm that information corresponds in the algorithm set is infused, it is candidate that markup information is used to characterize the sample determined using corresponding algorithm Whether value score is located at the corresponding value score section of sample product；

Using machine learning method, the corresponding sample candidate of sample attribute information in the sample attribute information aggregate is worth Score set is as input, using markup information corresponding with the sample candidate of input value score set as desired output, instruction Get order model.

9. according to the method described in claim 8, wherein, the markup information at least one markup information generated is right respectively The algorithm answered determines in accordance with the following steps in advance:

For the algorithm in the algorithm set, determine that the sample candidate according to determined by the algorithm is worth the accuracy rate of score；

Based on identified accuracy rate, selection algorithm is as algorithm corresponding with markup information from the algorithm set.

10. according to the method described in claim 8, wherein, the order model is obtained based on Random Forest model training Multi-tag disaggregated model.

11. a kind of method for pushed information, comprising:

Obtain the classification information and preset product information set of target user, wherein product information corresponds to predetermined Label, label are that method described in one of -10 generates according to claim 1；

Corresponding relationship based on preset, label and classification information, it is determining to believe with the classification from the product information set Cease matched product information；

Identified product information is pushed to the terminal of the target user.

12. a kind of for generating the device of information, comprising:

Acquiring unit is configured to obtain the attribute information of target product；

Determination unit is configured to determine candidate value score set based on the attribute information and preset algorithm set；

Generation unit is configured to the order model that the candidate value score set input is trained in advance, obtains probability value Set, wherein the probability value in the probability value set corresponds to the algorithm in the algorithm set, corresponds to for characterizing to utilize The order of accuarcy of candidate value score that determines of algorithm；

Storage unit is configured to the selection target probability value from the probability value set, and will be according to the destination probability It is worth label and storage that the candidate value score that corresponding algorithm determines is determined as the target product.

13. a kind of device for pushed information, comprising:

Acquiring unit is configured to obtain the classification information and preset product information set of target user, wherein product information Corresponding to predetermined label, label is that method described in one of -10 generates according to claim 1；

Determination unit is configured to the corresponding relationship based on preset, label and classification information, from the product information set In, the determining and matched product information of the classification information；

Push unit is configured to push to identified product information the terminal of the target user.

14. a kind of server, comprising:

One or more processors；

Storage device is stored thereon with 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 The now method as described in any in claim 1-11.

15. a kind of computer-readable medium, is stored thereon with computer program, wherein the realization when program is executed by processor Method as described in any in claim 1-11.