CN107590690A

CN107590690A - Data processing method, device and server

Info

Publication number: CN107590690A
Application number: CN201710790349.3A
Authority: CN
Inventors: 肖锦文; 赵嘉寅; 周琳
Original assignee: Alibaba Group Holding Ltd
Current assignee: Advanced New Technologies Co Ltd; Advantageous New Technologies Co Ltd
Priority date: 2017-09-05
Filing date: 2017-09-05
Publication date: 2018-01-16
Anticipated expiration: 2037-09-05
Also published as: CN107590690B

Abstract

This specification embodiment provides a kind of data processing method, device and server.This method includes：Original historical trading data is obtained, determines the characteristic vector of transaction shop and trade user in the historical trading data；Transaction feature vector using the characteristic vector of the trade user of the first predetermined number corresponding to target transaction shop as the target transaction shop；Transaction feature vector using the characteristic vector in the transaction shop of the second predetermined number corresponding to target transaction user as the target transaction user；The transaction representative learning of shop and trade user is traded using the transaction feature vector of target transaction user described in the transaction feature vector sum in target transaction shop, obtains the transaction insertion characteristic in corresponding trade user and shop of merchandising.

Description

Data processing method, device and server

Technical field

This specification embodiment is related to data-optimized processing technology field, more particularly to a kind of data processing method, device And server.

Background technology

With the development of internet Consumption Age, increasing people can buy commodity in some e-commerce platforms, be The buying rate of commodity is improved, often user and businessman can be inferred to based on the historical trading data in e-commerce platform Between the degree of association, and then recommend to businessman the user of the high degree of association, in order to which businessman targetedly carries out goods marketing, uses Safeguard at family；Or recommend the businessman of the high degree of association to user.

, often can be according to original historical trading number in the prior art it needs to be determined that the degree of association between businessman and user According to artificial setting and substantial amounts of feature is built, such as the trading volume of nearest 1 day, the dealing money of nearest one week, the moon transaction frequency Rate etc.；Then, the degree of association between businessman and user, such as user A are weighed in business according to the comparison between each feature Family's X dealing money of nearest one week is 280 yuan, and user B is 80 yuan in the businessman X dealing money of nearest one week, accordingly, according to This feature (dealing money of nearest one week), it can be determined that the degree of association gone out between user A and businessman X is more than user B and businessman X Between the degree of association.It is but above-mentioned to be used to determine that the feature of the degree of association between user and businessman needs to be manually set simultaneously in the prior art , easily there is invalid feature, and for substantial amounts of historical trading data, the problem for the treatment of effeciency is low often be present in structure.Cause This, it is desirable to provide more rapidly or more effective scheme.

The content of the invention

The purpose of this specification embodiment is to provide a kind of data processing method, device and server, can be with fast and effective Obtain characterize transaction feature transaction insertion characteristic, improve data-handling efficiency.

This specification embodiment is realized in：

A kind of data processing method, including：

Historical trading data is obtained, determines the characteristic vector of transaction shop and trade user in the historical trading data；

The characteristic vector of the trade user of the first predetermined number corresponding to target transaction shop is handed over as the target The transaction feature vector in easy shop；

The characteristic vector in the transaction shop of the second predetermined number corresponding to target transaction user is handed over as the target The transaction feature vector of easy user；

Handed over using the transaction feature vector of target transaction user described in the transaction feature vector sum in target transaction shop Easy shop and the transaction representative learning of trade user, obtain the transaction insertion characteristic in corresponding trade user and shop of merchandising According to.

A kind of data processing equipment, including：

Historical trading data acquisition module, for obtaining historical trading data；

Characteristic vector determining module, for determine to merchandise in the historical trading data shop and trade user feature to Amount；

First transaction feature vector determining module, for by the transaction of the first predetermined number corresponding to target transaction shop Transaction feature vector of the characteristic vector of user as the target transaction shop；

Second transaction feature vector determining module, for by the transaction of the second predetermined number corresponding to target transaction user Transaction feature vector of the characteristic vector in shop as the target transaction user；

Representative learning module, the friendship for target transaction user described in the transaction feature vector sum using target transaction shop Easy characteristic vector is traded the transaction representative learning of shop and trade user, obtains corresponding trade user and shop of merchandising The embedded characteristic of transaction.

A kind of data processing server, including processor and memory, the memory storage is by the computing device Computer program instructions, the computer program instructions include：

As seen from the above, this specification one or more embodiment is by obtaining original historical trading data；Then, directly Connect using the characteristic vector of the trade user of the first predetermined number corresponding to target transaction shop as the target transaction shop Transaction feature vector；Using the characteristic vector in the transaction shop of the second predetermined number corresponding to target transaction user as described in The transaction feature vector of target transaction user；Utilize target transaction user described in the transaction feature vector sum in target transaction shop Transaction feature vector is traded the transaction representative learning of shop and trade user, and transaction feature is increased into corresponding transaction uses In the characteristic vector in family and transaction shop, and then obtain characterizing the transaction insertion characteristic of transaction feature.This specification Embodiment, not only can be with directly using original transaction data study to the transaction insertion characteristic that can characterize transaction feature The transaction insertion characteristic learnt, which is effectively ensured, can directly characterize transaction feature, can also improve data-handling efficiency.

Brief description of the drawings

In order to illustrate more clearly of this specification one or more embodiment or technical scheme of the prior art, below will The required accompanying drawing used in embodiment or description of the prior art is briefly described, it should be apparent that, in describing below Accompanying drawing is only some embodiments described in this specification, for those of ordinary skill in the art, is not paying creation Property work on the premise of, other accompanying drawings can also be obtained according to these accompanying drawings.

Fig. 1 is a kind of schematic flow sheet of the embodiment for the data processing method that this specification provides；

Fig. 2 is target transaction user described in the transaction feature vector sum using target transaction shop of this specification offer Transaction feature vector is traded the transaction representative learning of shop and trade user, obtains corresponding trade user and transaction shop The embedded characteristic of transaction a kind of embodiment schematic flow sheet；

Fig. 3 is a kind of schematic flow sheet for embodiment that relationship characteristic data are hidden in the determination that this specification provides；

Fig. 4 carries out degree of association meter between user and businessman based on the data that the data processing method that this specification provides obtains A kind of schematic diagram for the embodiment calculated；

Fig. 5 is a kind of structural representation of the embodiment for the data processing equipment that this specification provides.

Embodiment

This specification embodiment provides a kind of data processing method, device and server.

In order that those skilled in the art more fully understand the technical scheme in this specification, below in conjunction with this explanation Accompanying drawing in book embodiment, the technical scheme in this specification embodiment is clearly and completely described, it is clear that described Embodiment be only this specification part of the embodiment, rather than whole embodiment.Based on the embodiment in this specification, The every other embodiment that those of ordinary skill in the art are obtained under the premise of creative work is not made, should all belong to The scope of this specification protection.

At present, on some e-commerce platforms, businessman generally requires to carry out goods marketing, user's maintenance etc. to improve business The buying rate of product.The record that historical trading data carries out shopping generation trading activity as user in businessman (shop), often may be used With for reflecting fancy grade of the user to shop.Therefore, correlated characteristic is often extracted to be used according to historical trading data The determination of the degree of association between family and shop, and then related recommendation process is carried out to improve the buying rate of commodity.Existing It is manually set and builds in substantial amounts of characteristic procedure according to original historical trading data, the feature artificially built often exists certain Subjectivity, easily there is invalid feature, cause the feature of extraction can not effectively to characterize transaction between user and shop Feature, (transaction feature can include the feature of the trading activity of reflection user).Based on this, can improve calculate user with Sign ability of the characteristic to transaction feature used by the degree of association between shop.

A kind of specific embodiment of this specification data processing method introduced below.Fig. 1 is the data that this specification provides A kind of schematic flow sheet of embodiment of processing method, present description provides the operation of the method as described in embodiment or flow chart Step, but either can include more or less operating procedures without performing creative labour based on conventional.Enumerated in embodiment The step of order be only numerous step execution sequences in a kind of mode, do not represent unique execution sequence.In practice , can be according to embodiment either method order execution shown in the drawings or parallel execution when system or client production perform (such as environment of parallel processor or multiple threads).Specifically as shown in figure 1, methods described can include：

S102：Historical trading data is obtained, determines the feature of transaction shop and trade user in the historical trading data Vector.

In this specification embodiment, substantial amounts of historical trading data can be obtained, and determine corresponding to historical trading data Each transaction shop and trade user characteristic vector.

Specifically, the historical trading data can include user and shop corresponding to exchange.The transaction shop Characteristic vector can include the random character vector of default dimension, and the value model of each characteristic value of random character vector Enclose for 0-1.Specifically, for example default dimension is three-dimensional, search random character vector can be (0.6,0.4,0.5).The friendship The random character that the characteristic vector of easy user can include default dimension is vectorial, and the default dimension of random character vector Characteristic value sum is 1.The characteristic vector in the transaction shop is identical with the dimension of the characteristic vector of the trade user.

S104：Using the characteristic vector of the trade user of the first predetermined number corresponding to target transaction shop as the mesh The transaction feature vector in mark transaction shop.

, can be by the spy of the trade user of the first predetermined number corresponding to target transaction shop in this specification embodiment Transaction feature vector of the sign vector as the target transaction shop.Specifically, in certain embodiments, the historical trading number According to exchange hour can also be included.Accordingly, the trade user of first predetermined number by corresponding to target transaction shop Characteristic vector can include as the transaction feature vector in the target transaction shop：

Exchange hour in transaction data according to corresponding to the target transaction shop is chosen and the target transaction shop Spread the characteristic vector of the trade user of the first corresponding predetermined number；

Transaction feature using the characteristic vector of the trade user of first predetermined number as the target transaction shop Vector.

In a specific embodiment, it is pre- that first corresponding with the target transaction shop is chosen according to exchange hour If the characteristic vector of the trade user of quantity can be included according to exchange hour successively, continuous or regular interval chooses the The characteristic vector of the trade user of one predetermined number；It can also include according to exchange hour, choose first in certain time period The characteristic vector of the trade user of predetermined number.Specifically, first predetermined number can be to be set previously according to actual demand The quantity put, such as 5 are arranged to, accordingly, the transaction feature vector in target transaction shop can be (U₁, U₂, U₃, U₄, U₅), its In, U₁, U₂, U₃, U₄, U₅For the characteristic vector of 5 trade users corresponding to the target transaction shop.

S106：Using the characteristic vector in the transaction shop of the second predetermined number corresponding to target transaction user as the mesh Mark the transaction feature vector of trade user.

, can be by the spy in the transaction shop of the second predetermined number corresponding to target transaction user in this specification embodiment Transaction feature vector of the sign vector as the target transaction user.Specifically, in certain embodiments, the historical trading number According to exchange hour can also be included.Accordingly, the transaction shop of second predetermined number by corresponding to target transaction user Characteristic vector can include as the transaction feature vector of the target transaction user：

Exchange hour in transaction data according to corresponding to the target transaction user is chosen to be used with the target transaction The characteristic vector in the transaction shop of the second corresponding predetermined number of family；

Transaction feature using the characteristic vector in the transaction shop of second predetermined number as the target transaction user Vector.

In a specific embodiment, it is pre- that second corresponding with the target transaction user is chosen according to exchange hour If the characteristic vector in the transaction shop of quantity can be included according to exchange hour successively, continuous or regular interval chooses the The characteristic vector in the transaction shop of two predetermined numbers；It can also include according to exchange hour, choose second in certain time period The characteristic vector in the transaction shop of predetermined number.Specifically, second predetermined number can be to be set previously according to actual demand The quantity put, second predetermined number can be different from first predetermined number, it is preferred that second predetermined number can With identical with first predetermined number.Specifically.Such as second predetermined number is arranged to 5, accordingly, target transaction is used The transaction feature vector at family can be (S₁, S₂, S₃, S₄, S₅), wherein, S₁, S₂, S₃, S₄, S₅It is right for the target transaction user The characteristic vector in the 5 transaction shops answered.

S108：Transaction feature vector using target transaction user described in the transaction feature vector sum in target transaction shop enters Row transaction shop and the transaction representative learning of trade user, obtain the transaction insertion feature in corresponding trade user and shop of merchandising Data.

Target transaction user described in the transaction feature vector sum in target transaction shop can be utilized in this specification embodiment Transaction feature vector be traded the transaction representative learning of shop and trade user, obtain corresponding trade user and transaction shop The transaction insertion characteristic of paving.As shown in Fig. 2 Fig. 2 is the transaction feature using target transaction shop that this specification provides The transaction feature vector of target transaction user described in vector sum is traded the transaction representative learning of shop and trade user, obtains The schematic flow sheet of a kind of embodiment of the embedded characteristic of transaction in corresponding trade user and transaction shop, specifically, can With including：

S202：Respectively to the transaction feature of target transaction user described in the transaction feature vector sum in the target transaction shop Vector is pre-processed, and the transaction for obtaining the target transaction shop and the target transaction user characterizes vector.

Specifically, in this specification embodiment, because transaction feature vector includes one or more characteristic vectors, in order to protect The characteristic vector that transaction feature is vectorial with corresponding characteristic vector is identical dimensional is demonstrate,proved, can be to handing in order to follow-up calculating Easy characteristic vector is pre-processed.Specifically, the pretreatment can include it is following any：

Mean value calculation processing, read group total processing, take maximum processing, weighted average calculating processing.

So that mean value calculation is handled as an example, average value can be carried out to multiple characteristic vectors corresponding to transaction feature vector Calculating is handled, the average value vector that then will be obtained, and vector is characterized as the transaction enjoyed great prestige.In addition, for transaction feature vector The situation of a corresponding characteristic vector, what the mean value calculation of a characteristic vector was handled to obtain is still this feature vector.

S204：The characteristic vector that transaction to the target transaction shop characterizes target transaction shop described in vector sum is carried out Normalized, the transaction calculated based on default loss function after the normalization of the target transaction shop characterize vector sum feature to Diversity factor between amount.

This specification embodiment can be to the target transaction shop transaction characterize vector sum described in target transaction shop Characteristic vector be normalized, when can so ensure counting loss function, between two vectors for identical dimensional and The vector of same order.Specifically, the normalized can include the side using softmax flexibility maximum transfer functions Formula, but this specification embodiment is not limited thereto.Transaction after corresponding normalization characterizes the feature of vector sum characteristic vector It is 1 to be worth sum, and the span of each characteristic value is in 0-1.A such as three-dimensional vector (0.1,0.3,0.6).

After normalized, the friendship after the target transaction shop normalization can be calculated based on default loss function Easily characterize the diversity factor between vector sum characteristic vector.Using the value of corresponding default loss function as diversity factor.Specifically, this Loss function is preset described in specification embodiment can at least include one of the following：

Cross Entropy Function, 0-1 loss functions, logarithm loss function, figure penalties function, negative cross entropy and difference of two squares letter Number.

By taking Cross Entropy Function as an example, H (S, Z)=∑ (Slog (S/Z)), here H (S, Z) represent that target transaction shop is returned Transaction after one change characterizes the relative entropy between vector and characteristic vector；S represent the feature after the normalization of target transaction shop to Amount；Transaction after the normalization of Z target transactions shop characterizes vector.The value for the relative entropy being calculated accordingly can be used as described Transaction after the normalization of target transaction shop characterizes diversity factor between vector and characteristic vector.

From above-mentioned, when the value corresponding to default loss function is bigger, the diversity factor between corresponding vector is bigger；Instead It, when the value corresponding to default loss function is smaller, the diversity factor between corresponding vector is smaller.

S206：The characteristic vector of the trade user of the first predetermined number corresponding to the target transaction shop is adjusted to change Become the transaction feature vector in the target transaction shop, repeat the pre- of the above-mentioned transaction feature vector to the target transaction shop Transaction of the processing to after being normalized the step of calculating diversity factor to the target transaction shop is characterized between vector sum characteristic vector Diversity factor meet the first preparatory condition.

Diversity factor between vector and characteristic vector is characterized in order to reduce the transaction after the normalization in the target transaction shop, The characteristic vector of the trade user of the first predetermined number corresponding to the target transaction shop can be adjusted to change the mesh The transaction feature vector in mark transaction shop, then repeats to enter the transaction feature vector in the target transaction shop in above-mentioned S202 Row pretreatment and the transaction after being normalized to the target transaction shop the step of S204 are characterized between vector sum characteristic vector Diversity factor meets the first preparatory condition.Wherein, first preparatory condition can include current diversity factor and last calculating It is pre- that the difference that the difference of obtained diversity factor is less than or equal between the first preset value or adjacent diversity factor is less than or equal to first If the number of value is more than or equal to the first preset times.Specifically, first preset value and first preset times can be tied Practical application request is closed to pre-set.

The diversity factor that transaction after the target transaction shop normalizes is characterized between vector sum characteristic vector meets the One preparatory condition, the transaction after the target transaction shop normalization characterize the diversity factor between the current characteristic vector of vector sum It is smaller and tend towards stability.

S208：By the feature of the trade user corresponding to target transaction shop when meeting first preparatory condition to The characteristic vector as the target transaction user is measured, the characteristic vector and transaction to the target transaction user characterize vector and entered Row normalized, the characteristic vector after the target transaction user normalization is calculated based on default loss function and transaction characterizes Diversity factor between vector.

Specifically, normalized here may refer to above-mentioned related step to the calculating based on default loss function Suddenly, will not be repeated here.

S210：The characteristic vector in the transaction shop of the second predetermined number corresponding to the target transaction user is adjusted to change Become the transaction feature vector of the target transaction user, repeat the pre- of the above-mentioned transaction feature vector to the target transaction user Transaction of the processing to after being normalized the step of calculating diversity factor to the target transaction user is characterized between vector sum characteristic vector Diversity factor meet the second preparatory condition.

Specifically, second preparatory condition can include current diversity factor and the last diversity factor being calculated The number that the difference that difference is less than or equal between the second preset value or adjacent diversity factor is less than or equal to the second preset value is more than Equal to the second preset times.Specifically, second preset value and second preset times can combine practical application request Pre-set.Second preset value can be identical with first preset value, can also be different；Second preset times can , can also be different with identical with first preset times.

The diversity factor that transaction after the target transaction user normalizes is characterized between vector sum characteristic vector meets the Two preparatory conditions, the transaction after the target transaction user normalization characterize the diversity factor between the current characteristic vector of vector sum It is smaller and tend towards stability.

S212：By the feature in the transaction shop corresponding to target transaction user when meeting second preparatory condition Characteristic vector of the vector as the target transaction shop, the step of above-mentioned pretreatment extremely calculates diversity factor is repeated to the target The diversity factor that transaction after the normalization of transaction shop is characterized between vector sum characteristic vector meets the first preparatory condition and the mesh The diversity factor that transaction after mark trade user normalization is characterized between vector sum characteristic vector meets the second preparatory condition.

S214：The diversity factor that transaction after the target transaction shop is normalized is characterized between vector sum characteristic vector accords with Close the diversity factor between the transaction sign vector sum characteristic vector after the first preparatory condition and target transaction user normalization The characteristic vector of trade user and shop of merchandising when meeting the second preparatory condition is respectively as corresponding trade user and transaction The transaction insertion characteristic in shop.

This specification embodiment repeats above-mentioned base by constantly alternately adjusting trade user and the characteristic vector in shop of merchandising The diversity factor between corresponding vector is calculated in default loss function, come can during reducing the calculated value of default loss function So that transaction feature to be increased in the characteristic vector in corresponding trade user and shop of merchandising, and then obtain that transaction spy can be characterized The transaction insertion characteristic of sign.

Furthermore, it is necessary to the step of illustrating, being enumerated in this specification embodiment is sequentially only that numerous steps execution are suitable A kind of mode in sequence, does not represent unique execution sequence.For example, first to target transaction shop in above-mentioned steps S204 and S206 Transaction characterize vector characteristic vector be normalized, corresponding diversity factor calculate and characteristic vector adjustment processing；So The characteristic vector of the transaction sign vector to target transaction user is normalized again afterwards, corresponding diversity factor calculates and feature The adjustment processing of vector.In further embodiments, can also first to target transaction user transaction characterize vector feature to Amount is normalized, corresponding diversity factor calculates and the adjustment of characteristic vector processing, then to the transaction in target transaction shop The characteristic vector of sign vector is normalized, corresponding diversity factor calculates and the adjustment of characteristic vector processing.Certainly, it is this In the case of, the characteristic vector in the transaction shop corresponding to target transaction user when meeting second preparatory condition can be made For the characteristic vector in the target transaction shop.

In further embodiments, in order to preferably characterize transaction feature, multiple loss functions can be used to generate multiple The embedded characteristic of transaction, to excavate different transaction features.Accordingly, methods described can also include：

Default loss function based at least two types calculates the characteristic vector of the target transaction user and the mesh The diversity factor between the transaction sign vector of trade user is marked, and calculates the tran list after the target transaction shop normalization Levy the diversity factor between the characteristic vector in target transaction shop described in vector sum；

Accordingly, the transaction insertion characteristic in the trade user and transaction shop comprises at least two kinds of transaction Embedded characteristic.

In further embodiments, as shown in figure 3, relationship characteristic data are hidden in the determination that Fig. 3, which is this specification, to be provided A kind of schematic flow sheet of embodiment.Specifically, it can include：

S302：Any two kinds of embedded characteristic of transaction in trade user and shop of merchandising is chosen respectively.

S304：Any two kinds of transaction of trade user is embedded in characteristic as input layer and output Layer, centre add the first default hiding layer building first nerves network.

Any embedded characteristic of two kinds of transaction can include a certain friendship respectively as input layer and output layer Easily embedded characteristic is as input layer, and another embedded characteristic of transaction is as output layer；And to input layer and output layer Transaction insertion characteristic be interchangeable.

Specifically, the described first default hidden layer can include one or more layers hidden layer.Building first god Can use tanh hyperbolic tangent functions through the described in during network first default activation primitive between hidden layer and input layer, but Specification embodiment is not limited thereto.

S306：Any two kinds of transaction in shop of merchandising is embedded in characteristic as input layer and output Layer, centre add the second default hiding layer building nervus opticus network.

Here the step of building nervus opticus network may refer to the description of above-mentioned structure first nerves network, herein no longer Repeat.

S308：The learning training of first nerves network based on structure, obtain any two kinds of friendship of trade user Hiding relationship characteristic data between easily embedded characteristic information.

S310：The learning training of nervus opticus network based on structure, obtain any two kinds of friendship in transaction shop Hiding relationship characteristic data between easily embedded characteristic information.

This specification embodiment can characterize transaction from different angles due to the embedded characteristic of different types of transaction Feature, this specification embodiment can be by obtaining hiding for the relation that can reflect between different types of transaction feature data Relationship characteristic data, it subsequently can preferably determine the degree of association between user and shop.

Due to the degree of association that the transaction between user and shop can reflect between user and shop, therefore, can utilize The embedded characteristic of transaction of transaction feature is characterized to determine the degree of association between user and shop, and enter based on the degree of association The corresponding recommendation process of row.Accordingly, methods described can also include：

Calculated using the embedded characteristic of transaction in merchandise shop and trade user between trade user and transaction shop The degree of association；

It is more than or equal to the transaction shop corresponding to the embedded characteristic of transaction of default relating value to the degree of association and transaction is used Recommendation process is carried out between family.

Specifically, the transaction insertion characteristic meter in merchandise shop and trade user is utilized described in this specification embodiment The transaction that the degree of association that trade user is calculated between shop of merchandising can include calculating the trade user and shop of merchandising is embedded in The distance between characteristic.In a specific embodiment, the distance between embedded characteristic of merchandising can be transaction Euclidean distance between embedded characteristic, when the numerical value for the Euclidean distance being calculated based on the embedded characteristic of two transaction is got over Small, the correlation degree that can represent corresponding trade user between shop of merchandising is better, and the degree of association is higher；It is based on conversely, working as The numerical value for the Euclidean distance that the embedded characteristic of two transaction is calculated is bigger, can represent corresponding trade user and transaction Correlation degree between shop is poorer, and the degree of association is lower.

Certainly, the distance between embedded characteristic of being merchandised in this specification embodiment be not limited only to it is above-mentioned it is European away from From, COS distance, manhatton distance etc. can also be included, between embedded characteristic of being merchandised described in this specification embodiment away from From not being limited with above-mentioned.

Specifically, the default relating value can include the recommendation condition pre-set according to practical application request.Work as friendship , can be to corresponding when the degree of association between the embedded characteristic of transaction in easy user and transaction shop is more than the default relating value Trade user and transaction shop between carry out recommendation process.

In further embodiments, methods described can also include：

The transaction in merchandise shop and trade user is embedded in characteristic respectively and hiding relationship characteristic data are made accordingly For corresponding transaction shop and the characterize data of trade user；

The degree of association for calculating trade user between shop of merchandising using the characterize data in merchandise shop and trade user；

The degree of association is more than or equal between the transaction shop corresponding to the characterize data of default relating value and trade user Row recommendation process.

Associating between trade user and shop of merchandising is calculated here with the characterize data of transaction shop and trade user Degree can include calculating the distance between characterize data of the trade user and transaction shop.Here between computational representation data Distance can be found in the distance between embedded characteristic of above-mentioned calculating transaction, will not be repeated here.

In further embodiments, methods described can also include：

Transaction insertion characteristic based on default machine learning algorithm to transaction shop and trade user, which is associated, to be pushed away Training is recommended, obtains the first correlation recommendation model；

At the recommendation that output result based on the first correlation recommendation model is traded user between shop of merchandising Reason.

This specification embodiment can be embedded in based on transaction of the default machine learning algorithm to transaction shop and trade user Characteristic is associated recommendation training, obtains the first correlation recommendation model, determines to hand over the first correlation recommendation model The probability of association between easy user and transaction shop, using the probability of association higher than default probable value as output result with Carry out recommendation process.Specifically, the default probable value can according in practical application to associating between user and shop The requirement of degree is set.

In further embodiments, methods described can also include：

Transaction based on default machine learning algorithm to transaction shop and trade user is embedded in characteristic and corresponding hidden Hide relationship characteristic data and be associated recommendation training, obtain the second correlation recommendation model；

At the recommendation that output result based on the second correlation recommendation model is traded user between shop of merchandising Reason.

This specification embodiment can be embedded in based on transaction of the default machine learning algorithm to transaction shop and trade user Characteristic and accordingly hiding relationship characteristic data are associated recommendation training, obtain the second correlation recommendation model, with described Second correlation recommendation model is come the probability of the association that determines trade user between shop of merchandising, by the probability of association higher than default Probable value conduct output result to carry out recommendation process.Specifically, the default probable value can be according to practical application In requirement to the degree of association between user and shop set.

As can be seen here, a kind of one or more embodiments of data processing method of this specification are by obtaining original history Transaction data；Then, directly using the characteristic vector of the trade user of the first predetermined number corresponding to target transaction shop as The transaction feature vector in the target transaction shop；By the transaction shop of the second predetermined number corresponding to target transaction user Transaction feature vector of the characteristic vector as the target transaction user；Utilize the transaction feature vector sum institute in target transaction shop The transaction feature vector for stating target transaction user is traded the transaction representative learning of shop and trade user, and transaction feature is increased It is added in the characteristic vector in corresponding trade user and shop of merchandising, and then the transaction insertion for obtaining characterizing transaction feature is special Levy data.This specification embodiment is directly embedded in special using original transaction data study to the transaction that can characterize transaction feature Data are levied, the transaction insertion characteristic learnt, which can not only be effectively ensured, can directly characterize transaction feature, can also improve Data-handling efficiency.

As shown in figure 4, the data that Fig. 4 is obtained based on the data processing method that this specification provides carry out user and businessman it Between calculation of relationship degree a kind of embodiment schematic diagram；The data provided below in conjunction with Fig. 4 introductions based on this specification embodiment The data that processing method obtains carry out a kind of embodiment of calculation of relationship degree between user and businessman, as can be seen from Fig. 4, at one In specific embodiment, according to historical trading data centered on shop, 5 users that transaction was carried out with the shop are chosen： User 1, and user 2, and user 3, user 4 and user 5；And according to historical trading data customer-centric, choose 5 and the use Family carried out the shop of transaction：Shop 1, shop 2, shop 3, shop 4 and shop 5；Then, to 5 users and the institute in 5 shops Corresponding characteristic vector carries out the pretreatment of mean value calculation；Then, by constantly alternately adjusting trade user and transaction shop Characteristic vector repeat normalized, and be based respectively on Cross Entropy Function peace variance function and calculate between corresponding vector Diversity factor, transaction feature can be increased to corresponding trade user during reducing the calculated value of default loss function With in the characteristic vector in shop of merchandising, and then obtaining characterizing transaction insertion characteristic U1, U2, S1, S2 of transaction feature； Then, determine to reflect the hiding relationship characteristic of relation between U1, U2 using embedded characteristic U1, U2 of transaction of user Data H_U12；And determine to reflect the hidden of relation between S1, S2 using embedded characteristic S1, S2 of transaction in shop Hide relationship characteristic data H_S12；Then, the characterize data using U1, U2, H_U12 as user, using S1, S2, H_S12 as shop The characterize data of paving；Finally, the pass between user and shop is determined based on the calculating between user and the characterize data in shop Connection degree.Here it is used for determining that the characteristic of the degree of association between user and shop not only includes that multiple can to characterize transaction special The transaction insertion characteristic of sign, in addition to the hiding relationship characteristic number of relation between the embedded characteristic of transaction can be reflected According to, realize that the feature for learning the various process based on original historical trading data carries out fusion, can be more preferable Embody the degree of association between user and shop.

On the other hand this specification also provides a kind of data processing equipment, Fig. 5 is the data processing dress that this specification provides The structural representation for a kind of embodiment put, as shown in figure 5, described device 500 can include：

Historical trading data acquisition module 510, it can be used for obtaining historical trading data；

Characteristic vector determining module 520, it is determined for merchandised in the historical trading data shop and trade user Characteristic vector；

First transaction feature vector determining module 530, it can be used for the first present count corresponding to target transaction shop Transaction feature vector of the characteristic vector of the trade user of amount as the target transaction shop；

Second transaction feature vector determining module 540, it can be used for the second present count corresponding to target transaction user Transaction feature vector of the characteristic vector in the transaction shop of amount as the target transaction user；

Representative learning module 550, it can be used for target transaction described in the transaction feature vector sum using target transaction shop The transaction feature vector of user is traded the transaction representative learning of shop and trade user, obtains corresponding trade user and friendship The transaction insertion characteristic in easy shop.

In another embodiment, the representative learning module 550 can include：

Pretreatment unit, it can be used for respectively to target transaction described in the transaction feature vector sum in the target transaction shop The transaction feature vector of user is pre-processed, and the transaction for obtaining the target transaction shop and the target transaction user characterizes Vector；

First normalized unit, it can be used for the transaction to the target transaction shop and characterize target described in vector sum The characteristic vector in transaction shop is normalized；

First diversity factor computing unit, it can be used for calculating the target transaction shop normalization based on default loss function Transaction afterwards characterizes the diversity factor between vector sum characteristic vector；

First adjustment processing unit, can be used for the friendship for adjusting the first predetermined number corresponding to the target transaction shop The characteristic vector of easy user is repeated above-mentioned to the target transaction shop with changing the transaction feature in target transaction shop vector Tran list of the pretreatment of the transaction feature vector of paving to after being normalized the step of calculating diversity factor to the target transaction shop Diversity factor between sign vector sum characteristic vector meets the first preparatory condition；

Second normalized unit, it can be used for meeting target transaction shop institute during first preparatory condition Characteristic vector of the characteristic vector of corresponding trade user as the target transaction user, to the spy of the target transaction user Sign vector sum transaction characterizes vector and is normalized；

Second diversity factor computing unit, it can be used for calculating the target transaction user normalization based on default loss function Characteristic vector and transaction afterwards characterizes the diversity factor between vector；

Second adjustment processing unit, can be used for the friendship for adjusting the second predetermined number corresponding to the target transaction user The characteristic vector in easy shop is repeated above-mentioned to target transaction use with changing the transaction feature of target transaction user vector Tran list of the pretreatment of the transaction feature vector at family to after being normalized the step of calculating diversity factor to the target transaction user Diversity factor between sign vector sum characteristic vector meets the second preparatory condition；

Data processing unit, it can be used for corresponding to target transaction user when meeting second preparatory condition Transaction shop characteristic vector of the characteristic vector as the target transaction shop, repeat above-mentioned pretreatment to calculating diversity factor The step of to the target transaction shop normalize after transaction characterize vector sum characteristic vector between diversity factor meet first The diversity factor that transaction after preparatory condition and target transaction user normalization is characterized between vector sum characteristic vector meets the Two preparatory conditions；

The embedded characteristic determining unit of transaction, it can be used for the transaction after the target transaction shop is normalized and characterize Diversity factor between vector sum characteristic vector meets the tran list after the first preparatory condition and target transaction user normalization The characteristic vector of trade user and shop of merchandising when diversity factor between sign vector sum characteristic vector meets the second preparatory condition Transaction respectively as corresponding trade user and transaction shop is embedded in characteristic.

In another embodiment, described device 500 can also include：

Data processing module, it can be used for the default loss function based at least two types and calculate the target transaction use The transaction of the characteristic vector at family and the target transaction user characterize the diversity factor between vector, and calculate the target transaction Transaction after the normalization of shop characterizes the diversity factor between the characteristic vector in target transaction shop described in vector sum；

In another embodiment, described device 500 can also include：

The embedded characteristic of transaction chooses module, can be used for choose trade user and shop of merchandising respectively any two kinds The transaction insertion characteristic of type；

First nerves network struction module, it can be used for any two kinds of embedded characteristic of transaction of trade user According to respectively as input layer and output layer, centre adds the first default hiding layer building first nerves network；

Nervus opticus network struction module, it can be used for any two kinds of embedded characteristic of transaction in shop of merchandising According to respectively as input layer and output layer, centre adds the second default hiding layer building nervus opticus network；

It first learning training module, can be used for the learning training of the first nerves network based on structure, obtain transaction and use Hiding relationship characteristic data between any two kinds of embedded characteristic information of transaction at family；

Second learning training module, it can be used for the learning training of the nervus opticus network based on structure, obtain transaction shop Hiding relationship characteristic data between any two kinds of embedded characteristic information of transaction of paving.

In another embodiment, the pretreatment can include following any：

In another embodiment, the historical trading data can also include：Exchange hour.

In another embodiment, the first transaction feature vector determining module 530 can include：

First eigenvector chooses unit, can be used in the transaction data according to corresponding to the target transaction shop Exchange hour chooses the characteristic vector of the trade user of first predetermined number corresponding with the target transaction shop；

First transaction feature vector determination unit, can be used for by the feature of the trade user of first predetermined number to Measure the transaction feature vector as the target transaction shop.

In another embodiment, the second transaction feature vector determining module 540 can include：

Second feature vector chooses unit, can be used in the transaction data according to corresponding to the target transaction user Exchange hour chooses the characteristic vector in the transaction shop of second predetermined number corresponding with the target transaction user；

Second transaction feature vector determination unit, can be used for by second predetermined number transaction shop feature to Measure the transaction feature vector as the target transaction user.

In another embodiment, described device can also include：

First calculation of relationship degree module, it can be used for the transaction insertion characteristic meter using merchandise shop and trade user The degree of association that trade user is calculated between shop of merchandising；

First recommendation process module, the transaction that can be used for being more than or equal to the degree of association default relating value are embedded in characteristic Recommendation process is carried out between corresponding transaction shop and trade user.

In another embodiment, described device 500 can also include：

Characterize data determining module, can be used for respectively by the transaction of merchandise shop and trade user be embedded in characteristic and Corresponding hiding relationship characteristic data are as corresponding transaction shop and the characterize data of trade user；

Second calculation of relationship degree module, it can be used for calculating transaction use using the characterize data in merchandise shop and trade user The degree of association between family and transaction shop；

Second recommendation process module, it can be used for being more than or equal to the degree of association corresponding to the characterize data of default relating value Recommendation process is carried out between transaction shop and trade user.

In another embodiment, described device 500 can also include：

First correlation recommendation model determining module, it can be used for based on default machine learning algorithm to transaction shop and transaction The transaction insertion characteristic of user is associated recommendation training, obtains the first correlation recommendation model；

3rd recommendation process module, it can be used for the output result based on the first correlation recommendation model and be traded use Recommendation process between family and transaction shop.

In another embodiment, described device 500 can also include：

Second correlation recommendation model determining module, it can be used for based on default machine learning algorithm to transaction shop and transaction The transaction insertion characteristic of user and accordingly hiding relationship characteristic data, which are associated, recommends training, obtains the second association and pushes away Recommend model；

4th recommendation process module, it can be used for the output result based on the second correlation recommendation model and be traded use Recommendation process between family and transaction shop.

The above-mentioned data processing method or device that this specification embodiment provides can be in a computer by computing devices Corresponding programmed instruction is realized, is such as realized using the c++ language of windows operating systems at PC ends, or other for example using Android, iOS system programming language are realized in intelligent terminal, and processing logic realization based on quantum computer etc.. Therefore, on the other hand this specification also provides a kind of data processing server, including processor and memory, and the memory is deposited The computer program instructions by the computing device are stored up, the computer program instructions can include：

Specifically, in this specification embodiment, described processor can include central processing unit (CPU), also may be used certainly With including other single-chip microcomputers with logic processing capability, logic gates, integrated circuit etc., or its is appropriately combined.It is described Memory can be including nonvolatile memory etc..

The transaction feature vector of target transaction user enters described in the transaction feature vector sum using target transaction shop Row transaction shop and the transaction representative learning of trade user, obtain the transaction insertion feature in corresponding trade user and shop of merchandising Data can include：

Respectively to the transaction feature vector of target transaction user described in the transaction feature vector sum in the target transaction shop Pre-processed, the transaction for obtaining the target transaction shop and the target transaction user characterizes vector；

The characteristic vector that transaction to the target transaction shop characterizes target transaction shop described in vector sum carries out normalizing Change is handled, the transaction calculated based on default loss function after the normalization of the target transaction shop characterize vector sum characteristic vector it Between diversity factor；

The characteristic vector of the trade user of the first predetermined number corresponding to the target transaction shop is adjusted to change The transaction feature vector in target transaction shop is stated, repeats the pretreatment of the above-mentioned transaction feature vector to the target transaction shop Transaction to after being normalized the step of calculating diversity factor to the target transaction shop characterizes the difference between vector sum characteristic vector Different degree meets the first preparatory condition；

The characteristic vector of trade user corresponding to target transaction shop when meeting first preparatory condition is made For the characteristic vector of the target transaction user, characteristic vector and transaction to the target transaction user characterize vector and returned One change is handled, and calculates the characteristic vector after the target transaction user normalization based on default loss function and transaction characterizes vector Between diversity factor；

The characteristic vector in transaction shop of the second predetermined number corresponding to the target transaction user is adjusted to change The transaction feature vector of target transaction user is stated, repeats the pretreatment of the above-mentioned transaction feature vector to the target transaction user Transaction to after being normalized the step of calculating diversity factor to the target transaction user characterizes the difference between vector sum characteristic vector Different degree meets the second preparatory condition；

By the characteristic vector in the transaction shop corresponding to target transaction user when meeting second preparatory condition As the characteristic vector in the target transaction shop, the step of above-mentioned pretreatment extremely calculates diversity factor is repeated to the target transaction The diversity factor that transaction after the normalization of shop is characterized between vector sum characteristic vector meets the first preparatory condition and the target is handed over The diversity factor that transaction after easy user's normalization is characterized between vector sum characteristic vector meets the second preparatory condition；

The diversity factor that transaction after the target transaction shop is normalized is characterized between vector sum characteristic vector meets the The diversity factor that transaction after one preparatory condition and target transaction user normalization is characterized between vector sum characteristic vector meets The characteristic vector of trade user and transaction shop during the second preparatory condition is respectively as corresponding trade user and transaction shop Transaction insertion characteristic.

In another embodiment, the computer program instructions can also include：

Any two kinds of embedded characteristic of transaction in trade user and shop of merchandising is chosen respectively；

It is middle using any two kinds of embedded characteristic of transaction of trade user as input layer and output layer Add the first default hiding layer building first nerves network；

It is middle using any two kinds of embedded characteristic of transaction in shop of merchandising as input layer and output layer Add the second default hiding layer building nervus opticus network；

The learning training of first nerves network based on structure, obtain any two kinds of transaction insertion of trade user Hiding relationship characteristic data between characteristic information；

The learning training of nervus opticus network based on structure, obtain any two kinds of transaction insertion in transaction shop Hiding relationship characteristic data between characteristic information.

In another embodiment, wherein, the pretreatment can include following any：

In another embodiment, wherein, the historical trading data also includes：Exchange hour.

In another embodiment, the feature of the trade user of first predetermined number by corresponding to target transaction shop Vector includes as the transaction feature vector in the target transaction shop：

In another embodiment, second predetermined number by corresponding to target transaction user transaction shop feature to Measure includes as the transaction feature vector of the target transaction user：

In another embodiment, the computer program instructions also include：

In another embodiment, the computer program instructions can also include：

As can be seen here, the embodiment of a kind of data processing method of this specification, device or server is original by obtaining Historical trading data；Then, directly by the characteristic vector of the trade user of the first predetermined number corresponding to target transaction shop Transaction feature vector as the target transaction shop；By the transaction shop of the second predetermined number corresponding to target transaction user Transaction feature vector of the characteristic vector of paving as the target transaction user；Utilize the transaction feature vector in target transaction shop The transaction representative learning of shop and trade user is traded with the transaction feature vector of the target transaction user, transaction is special Sign increases in the characteristic vector in corresponding trade user and shop of merchandising, and then the transaction for obtaining characterizing transaction feature is embedding Enter characteristic.This specification embodiment is directly embedding to the transaction that can characterize transaction feature using original transaction data study Enter characteristic, the transaction insertion characteristic learnt, which can not only be effectively ensured, can directly characterize transaction feature, can be with Improve data-handling efficiency.

It is above-mentioned that this specification specific embodiment is described.Other embodiments are in the scope of the appended claims It is interior.In some cases, the action recorded in detail in the claims or step can be come according to different from the order in embodiment Perform and still can realize desired result.In addition, the process described in the accompanying drawings not necessarily require show it is specific suitable Sequence or consecutive order could realize desired result.In some embodiments, multitasking and parallel processing be also can With or be probably favourable.

In the 1990s, the improvement for a technology can clearly distinguish be on hardware improvement (for example, Improvement to circuit structures such as diode, transistor, switches) or software on improvement (improvement for method flow).So And as the development of technology, the improvement of current many method flows can be considered as directly improving for hardware circuit. Designer nearly all obtains corresponding hardware circuit by the way that improved method flow is programmed into hardware circuit.Cause This, it cannot be said that the improvement of a method flow cannot be realized with hardware entities module.For example, PLD (Programmable Logic Device, PLD) (such as field programmable gate array (Field Programmable Gate Array, FPGA)) it is exactly such a integrated circuit, its logic function is determined by user to device programming.By designer Voluntarily programming comes a digital display circuit " integrated " on a piece of PLD, without asking chip maker to design and make Special IC chip.Moreover, nowadays, substitution manually makes IC chip, this programming is also used instead mostly " patrols Volume compiler (logic compiler) " software realizes that software compiler used is similar when it writes with program development, And the source code before compiling also write by handy specific programming language, this is referred to as hardware description language (Hardware Description Language, HDL), and HDL is also not only a kind of, but have many kinds, such as ABEL (Advanced Boolean Expression Language)、AHDL(Altera Hardware Description Language)、Confluence、CUPL(Cornell University Programming Language)、HDCal、JHDL (Java Hardware Description Language)、Lava、Lola、MyHDL、PALASM、RHDL(Ruby Hardware Description Language) etc., VHDL (Very-High-Speed are most generally used at present Integrated Circuit Hardware Description Language) and Verilog.Those skilled in the art also should This understands, it is only necessary to method flow slightly programming in logic and is programmed into integrated circuit with above-mentioned several hardware description languages, Can is readily available the hardware circuit for realizing the logical method flow.

Controller can be implemented in any suitable manner, for example, controller can take such as microprocessor or processing Device and storage can by the computer of the computer readable program code (such as software or firmware) of (micro-) computing device Read medium, gate, switch, application specific integrated circuit (Application Specific Integrated Circuit, ASIC), the form of programmable logic controller (PLC) and embedded microcontroller, the example of controller include but is not limited to following microcontroller Device：ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20 and Silicone Labs C8051F320, are deposited Memory controller is also implemented as a part for the control logic of memory.It is also known in the art that except with Pure computer readable program code mode realized beyond controller, completely can be by the way that method and step is carried out into programming in logic to make Controller is obtained in the form of gate, switch, application specific integrated circuit, programmable logic controller (PLC) and embedded microcontroller etc. to come in fact Existing identical function.Therefore this controller is considered a kind of hardware component, and various for realizing to including in it The device of function can also be considered as the structure in hardware component.Or even, can be by for realizing that the device of various functions regards For that not only can be the software module of implementation method but also can be the structure in hardware component.

Device, module or the unit that above-described embodiment illustrates, it can specifically be realized by computer chip or entity, Huo Zheyou Product with certain function is realized.One kind typically realizes that equipment is computer.Specifically, computer for example can be individual People's computer, laptop computer, cell phone, camera phone, smart phone, personal digital assistant, media player, navigation Any equipment in equipment, electronic mail equipment, game console, tablet PC, wearable device or these equipment Combination.

For convenience of description, it is divided into various units during description apparatus above with function to describe respectively.Certainly, this is being implemented The function of each unit can be realized in same or multiple softwares and/or hardware during specification.

It should be understood by those skilled in the art that, embodiments of the invention can be provided as method, apparatus or computer program Product.Therefore, the present invention can use the reality in terms of complete hardware embodiment, complete software embodiment or combination software and hardware Apply the form of example.Moreover, the present invention can use the computer for wherein including computer usable program code in one or more The computer program production that usable storage medium is implemented on (including but is not limited to magnetic disk storage, CD-ROM, optical memory etc.) The form of product.

The present invention is the flow with reference to method according to embodiments of the present invention, equipment (device) and computer program product Figure and/or block diagram describe.It should be understood that can be by every first-class in computer program instructions implementation process figure and/or block diagram Journey and/or the flow in square frame and flow chart and/or block diagram and/or the combination of square frame.These computer programs can be provided The processors of all-purpose computer, special-purpose computer, Embedded Processor or other programmable data processing devices is instructed to produce A raw machine so that produced by the instruction of computer or the computing device of other programmable data processing devices for real The device for the function of being specified in present one flow of flow chart or one square frame of multiple flows and/or block diagram or multiple square frames.

These computer program instructions, which may be alternatively stored in, can guide computer or other programmable data processing devices with spy Determine in the computer-readable memory that mode works so that the instruction being stored in the computer-readable memory, which produces, to be included referring to Make the manufacture of device, the command device realize in one flow of flow chart or multiple flows and/or one square frame of block diagram or The function of being specified in multiple square frames.

These computer program instructions can be also loaded into computer or other programmable data processing devices so that counted Series of operation steps is performed on calculation machine or other programmable devices to produce computer implemented processing, so as in computer or The instruction performed on other programmable devices is provided for realizing in one flow of flow chart or multiple flows and/or block diagram one The step of function of being specified in individual square frame or multiple square frames.

In a typical configuration, computing device includes one or more processors (CPU), input/output interface, net Network interface and internal memory.

Internal memory may include computer-readable medium in volatile memory, random access memory (RAM) and/or The forms such as Nonvolatile memory, such as read-only storage (ROM) or flash memory (flash RAM).Internal memory is computer-readable medium Example.

Computer-readable medium includes permanent and non-permanent, removable and non-removable media can be by any method Or technology come realize information store.Information can be computer-readable instruction, data structure, the module of program or other data. The example of the storage medium of computer includes, but are not limited to phase transition internal memory (PRAM), static RAM (SRAM), moved State random access memory (DRAM), other kinds of random access memory (RAM), read-only storage (ROM), electric erasable Programmable read only memory (EEPROM), fast flash memory bank or other memory techniques, read-only optical disc read-only storage (CD-ROM), Digital versatile disc (DVD) or other optical storages, magnetic cassette tape, the storage of tape magnetic rigid disk, graphene stores or other Magnetic storage apparatus or any other non-transmission medium, the information that can be accessed by a computing device available for storage.According to herein In define, computer-readable medium does not include the data of temporary computer readable media (transitory media), such as modulation Signal and carrier wave.

It should also be noted that, term " comprising ", "comprising" or its any other variant are intended to nonexcludability Comprising so that process, method, commodity or equipment including a series of elements not only include those key elements, but also wrapping Include the other element being not expressly set out, or also include for this process, method, commodity or equipment intrinsic want Element.In the absence of more restrictions, the key element limited by sentence "including a ...", it is not excluded that wanted including described Other identical element also be present in the process of element, method, commodity or equipment.

It will be understood by those skilled in the art that the embodiment of this specification can be provided as method, apparatus or computer program production Product.Therefore, this specification can use the implementation in terms of complete hardware embodiment, complete software embodiment or combination software and hardware The form of example.Moreover, this specification can use the computer for wherein including computer usable program code in one or more The computer program production that usable storage medium is implemented on (including but is not limited to magnetic disk storage, CD-ROM, optical memory etc.) The form of product.

This specification can be described in the general context of computer executable instructions, such as journey Sequence module.Usually, program module include performing particular task or realize the routine of particular abstract data type, program, object, Component, data structure etc..This specification can also be put into practice in a distributed computing environment, in these DCEs In, by performing task by communication network and connected remote processing devices.In a distributed computing environment, program module It can be located in the local and remote computer-readable storage medium including storage device.

Each embodiment in this specification is described by the way of progressive, identical similar portion between each embodiment Divide mutually referring to what each embodiment stressed is the difference with other embodiment.Especially for device and For server example, because it is substantially similar to embodiment of the method, so description is fairly simple, related part is referring to side The part explanation of method embodiment.

The embodiment of this specification is the foregoing is only, is not limited to this specification.For art technology For personnel, this specification can have various modifications and variations.It is all this specification spirit and principle within made it is any Modification, equivalent substitution, improvement etc., should be included within right.

Claims

1. a kind of data processing method, including：

Using the characteristic vector of the trade user of the first predetermined number corresponding to target transaction shop as the target transaction shop The transaction feature vector of paving；

The characteristic vector in the transaction shop of the second predetermined number corresponding to target transaction user is used as the target transaction The transaction feature vector at family；

Shop is traded using the transaction feature vector of target transaction user described in the transaction feature vector sum in target transaction shop Paving and the transaction representative learning of trade user, obtain the transaction insertion characteristic in corresponding trade user and shop of merchandising.

2. the method according to claim 11, wherein, mesh described in the transaction feature vector sum using target transaction shop The transaction feature vector of mark trade user is traded the transaction representative learning of shop and trade user, obtains corresponding transaction and uses The transaction insertion characteristic in family and transaction shop includes：

The transaction feature vector of target transaction user described in the transaction feature vector sum in the target transaction shop is carried out respectively Pretreatment, the transaction for obtaining the target transaction shop and the target transaction user characterize vector；

Place is normalized in the characteristic vector that transaction to the target transaction shop characterizes target transaction shop described in vector sum Reason, the transaction calculated based on default loss function after the target transaction shop normalization are characterized between vector sum characteristic vector Diversity factor；

The characteristic vector of the trade user of the first predetermined number corresponding to the target transaction shop is adjusted to change the mesh The transaction feature vector in mark transaction shop, the pretreatment for repeating the above-mentioned transaction feature vector to the target transaction shop are extremely counted Transaction after the step of calculating diversity factor to target transaction shop normalization characterizes the diversity factor between vector sum characteristic vector Meet the first preparatory condition；

Using the characteristic vector of the trade user corresponding to target transaction shop when meeting first preparatory condition as institute The characteristic vector of target transaction user is stated, the characteristic vector and transaction to the target transaction user characterize vector and be normalized Processing, calculated based on default loss function between the characteristic vector after the target transaction user normalization and transaction sign vector Diversity factor；

The characteristic vector in the transaction shop of the second predetermined number corresponding to the target transaction user is adjusted to change the mesh The transaction feature vector of trade user is marked, the pretreatment for repeating the above-mentioned transaction feature vector to the target transaction user is extremely counted Transaction after the step of calculating diversity factor to target transaction user normalization characterizes the diversity factor between vector sum characteristic vector Meet the second preparatory condition；

Using corresponding to target transaction user when meeting second preparatory condition transaction shop characteristic vector as The characteristic vector in the target transaction shop, the step of above-mentioned pretreatment extremely calculates diversity factor is repeated to the target transaction shop The diversity factor that transaction after normalization is characterized between vector sum characteristic vector meets the first preparatory condition and the target transaction is used The diversity factor that transaction after the normalization of family is characterized between vector sum characteristic vector meets the second preparatory condition；

It is pre- that the diversity factor that transaction after the target transaction shop is normalized is characterized between vector sum characteristic vector meets first If the diversity factor that the transaction after condition and target transaction user normalization is characterized between vector sum characteristic vector meets second Friendship of the characteristic vector of trade user and transaction shop during preparatory condition respectively as corresponding trade user and transaction shop Easily embedded characteristic.

3. according to the method for claim 2, wherein, methods described also includes：

Default loss function based at least two types calculates the characteristic vector of the target transaction user and the target is handed over The transaction of easy user characterizes the diversity factor between vector, and the transaction calculated after the target transaction shop normalization characterize to Diversity factor between amount and the characteristic vector in the target transaction shop；

Accordingly, the transaction insertion characteristic in the trade user and transaction shop comprises at least two kinds of transaction and is embedded in Characteristic.

4. according to the method for claim 3, wherein, methods described also includes：

Any two kinds of embedded characteristic of transaction of trade user is added as input layer and output layer, centre First default hiding layer building first nerves network；

Any two kinds of embedded characteristic of transaction in shop of merchandising is added as input layer and output layer, centre Second default hiding layer building nervus opticus network；

The learning training of first nerves network based on structure, obtain any two kinds of embedded feature of transaction of trade user Hiding relationship characteristic data between information；

The learning training of nervus opticus network based on structure, obtain any two kinds of embedded feature of transaction in transaction shop Hiding relationship characteristic data between information.

5. according to the method described in claim 2 to 4 any one, wherein, the pretreatment includes following any：

6. according to the method described in Claims 1-4 any one, wherein, the historical trading data also includes：During transaction Between.

7. the method according to claim 11, wherein, the friendship of first predetermined number by corresponding to target transaction shop The characteristic vector of easy user includes as the transaction feature vector in the target transaction shop：

Exchange hour in transaction data according to corresponding to the target transaction shop is chosen and target transaction shop phase The characteristic vector of the trade user of corresponding first predetermined number；

Transaction feature vector using the characteristic vector of the trade user of first predetermined number as the target transaction shop.

8. the method according to claim 11, wherein, the friendship of second predetermined number by corresponding to target transaction user The characteristic vector in easy shop includes as the transaction feature vector of the target transaction user：

Exchange hour in transaction data according to corresponding to the target transaction user is chosen and the target transaction user phase The characteristic vector in the transaction shop of corresponding second predetermined number；

Transaction feature vector using the characteristic vector in the transaction shop of second predetermined number as the target transaction user.

9. according to the method described in Claims 1-4 any one, wherein, methods described also includes：

Associating between trade user and shop of merchandising is calculated using the embedded characteristic of transaction in merchandise shop and trade user Degree；

The degree of association is more than or equal to the transaction shop corresponding to the embedded characteristic of transaction of default relating value and trade user it Between carry out recommendation process.

10. according to the method for claim 4, wherein, methods described also includes：

The transaction in merchandise shop and trade user is embedded in characteristic respectively and hides relationship characteristic data accordingly as phase The transaction shop answered and the characterize data of trade user；

The degree of association is more than or equal between the transaction shop corresponding to the characterize data of default relating value and trade user and pushed away Recommend processing.

11. according to the method described in Claims 1-4 any one, wherein, methods described also includes：

Transaction insertion characteristic based on default machine learning algorithm to transaction shop and trade user is associated recommendation instruction Practice, obtain the first correlation recommendation model；

The recommendation process that output result based on the first correlation recommendation model is traded user between shop of merchandising.

12. according to the method for claim 4, wherein, methods described also includes：

Transaction based on default machine learning algorithm to transaction shop and trade user is embedded in characteristic and corresponding hide is closed It is that characteristic is associated recommendation training, obtains the second correlation recommendation model；

The recommendation process that output result based on the second correlation recommendation model is traded user between shop of merchandising.

13. a kind of data processing equipment, including：

Characteristic vector determining module, for determining the characteristic vector of transaction shop and trade user in the historical trading data；

First transaction feature vector determining module, for by the trade user of the first predetermined number corresponding to target transaction shop Characteristic vector as the target transaction shop transaction feature vector；

Second transaction feature vector determining module, for by the transaction shop of the second predetermined number corresponding to target transaction user Characteristic vector as the target transaction user transaction feature vector；

Representative learning module, the transaction for target transaction user described in the transaction feature vector sum using target transaction shop are special Sign vector is traded the transaction representative learning of shop and trade user, obtains the transaction in corresponding trade user and shop of merchandising Embedded characteristic.

14. device according to claim 13, wherein, the representative learning module includes：

Pretreatment unit, for the friendship to target transaction user described in the transaction feature vector sum in the target transaction shop respectively Easy characteristic vector is pre-processed, and the transaction for obtaining the target transaction shop and the target transaction user characterizes vector；

First normalized unit, target transaction shop described in vector sum is characterized for the transaction to the target transaction shop Characteristic vector be normalized；

First diversity factor computing unit, for calculating the transaction after the target transaction shop normalization based on default loss function Characterize the diversity factor between vector sum characteristic vector；

First adjustment processing unit, for adjusting the trade user of the first predetermined number corresponding to the target transaction shop Characteristic vector repeats the above-mentioned transaction to the target transaction shop to change the transaction feature in target transaction shop vector The pretreatment of characteristic vector to the transaction after being normalized the step of calculating diversity factor to the target transaction shop characterizes vector sum Diversity factor between characteristic vector meets the first preparatory condition；

Second normalized unit, for the friendship corresponding to by target transaction shop when meeting first preparatory condition Characteristic vector of the characteristic vector of easy user as the target transaction user, characteristic vector to the target transaction user and Transaction characterizes vector and is normalized；

Second diversity factor computing unit, for calculating the feature after the target transaction user normalization based on default loss function Vector sum transaction characterizes the diversity factor between vector；

Second adjustment processing unit, for adjusting the transaction shop of the second predetermined number corresponding to the target transaction user Characteristic vector repeats the above-mentioned transaction to the target transaction user to change the transaction feature of target transaction user vector The pretreatment of characteristic vector to the transaction after being normalized the step of calculating diversity factor to the target transaction user characterizes vector sum Diversity factor between characteristic vector meets the second preparatory condition；

Data processing unit, for the transaction shop corresponding to by target transaction user when meeting second preparatory condition Characteristic vector of the characteristic vector of paving as the target transaction shop, repeat the step of above-mentioned pretreatment extremely calculates diversity factor extremely The diversity factor that transaction after the target transaction shop normalization is characterized between vector sum characteristic vector meets the first preparatory condition And the diversity factor that the transaction after the target transaction user normalization is characterized between vector sum characteristic vector meets the second default bar Part；

The embedded characteristic determining unit of transaction, it is special to characterize vector sum for the transaction after the target transaction shop is normalized The transaction that diversity factor between sign vector meets after the first preparatory condition and target transaction user normalization characterizes vector sum The characteristic vector of trade user and transaction shop when diversity factor between characteristic vector meets the second preparatory condition respectively as Corresponding trade user and the transaction insertion characteristic in transaction shop.

15. device according to claim 14, wherein, described device also includes：

Data processing module, the feature of the target transaction user is calculated for the default loss function based at least two types The transaction of target transaction user described in vector sum characterizes the diversity factor between vector, and calculates target transaction shop normalizing Transaction after change characterizes the diversity factor between the characteristic vector in target transaction shop described in vector sum；

16. device according to claim 15, wherein, described device also includes：

The embedded characteristic of transaction chooses module, for choosing any two kinds of friendship of trade user and shop of merchandising respectively Easily embedded characteristic；

First nerves network struction module, for any two kinds of embedded characteristic of transaction of trade user to be made respectively For input layer and output layer, centre adds the first default hiding layer building first nerves network；

Nervus opticus network struction module, for any two kinds of embedded characteristic of transaction in shop of merchandising to be made respectively For input layer and output layer, centre adds the second default hiding layer building nervus opticus network；

First learning training module, for the learning training of the first nerves network based on structure, obtain any of trade user Hiding relationship characteristic data between the two kinds of embedded characteristic information of transaction；

Second learning training module, for the learning training of the nervus opticus network based on structure, obtain any of shop of merchandising Hiding relationship characteristic data between the two kinds of embedded characteristic information of transaction.

17. according to the device described in claim 14 to 16 any one, wherein, the pretreatment includes following any：

18. according to the device described in claim 14 to 16 any one, wherein, the historical trading data also includes：Transaction Time.

19. device according to claim 18, wherein, the first transaction feature vector determining module includes：

First eigenvector chooses unit, for the exchange hour in the transaction data according to corresponding to the target transaction shop Choose the characteristic vector of the trade user of first predetermined number corresponding with the target transaction shop；

First transaction feature vector determination unit, for using the characteristic vector of the trade user of first predetermined number as institute State the transaction feature vector in target transaction shop.

20. device according to claim 18, wherein, the second transaction feature vector determining module includes：

Second feature vector chooses unit, for the exchange hour in the transaction data according to corresponding to the target transaction user Choose the characteristic vector in the transaction shop of second predetermined number corresponding with the target transaction user；

Second transaction feature vector determination unit, for using the characteristic vector in the transaction shop of second predetermined number as institute State the transaction feature vector of target transaction user.

21. according to the device described in claim 13 to 16 any one, wherein, described device also includes：

First calculation of relationship degree module, used for calculating transaction using the transaction insertion characteristic in merchandise shop and trade user The degree of association between family and transaction shop；

First recommendation process module, for being more than or equal to the degree of association corresponding to the transaction insertion characteristic of default relating value Recommendation process is carried out between transaction shop and trade user.

22. device according to claim 16, wherein, described device also includes：

Characterize data determining module, for the transaction of merchandise shop and trade user to be embedded in into characteristic and corresponding hidden respectively Relationship characteristic data are hidden as corresponding transaction shop and the characterize data of trade user；

Second calculation of relationship degree module, for calculating trade user and transaction using the characterize data in merchandise shop and trade user The degree of association between shop；

Second recommendation process module, for being more than or equal to the transaction shop corresponding to the characterize data of default relating value to the degree of association Recommendation process is carried out between trade user.

23. according to the device described in claim 13 to 16 any one, wherein, described device also includes：

First correlation recommendation model determining module, for based on friendship of the default machine learning algorithm to transaction shop and trade user Easily embedded characteristic is associated recommendation training, obtains the first correlation recommendation model；

3rd recommendation process module, user and transaction are traded for the output result based on the first correlation recommendation model Recommendation process between shop.

24. device according to claim 16, wherein, described device also includes：

Second correlation recommendation model determining module, for based on friendship of the default machine learning algorithm to transaction shop and trade user Easily embedded characteristic and corresponding hiding relationship characteristic data are associated recommendation training, obtain the second correlation recommendation model；

4th recommendation process module, user and transaction are traded for the output result based on the second correlation recommendation model Recommendation process between shop.

25. a kind of data processing server, including processor and memory, the memory storage is by the computing device Computer program instructions, the computer program instructions include：

26. server according to claim 25, wherein, the transaction feature vector sum institute using target transaction shop The transaction feature vector for stating target transaction user is traded the transaction representative learning of shop and trade user, is handed over accordingly The transaction insertion characteristic in easy user and transaction shop includes：

27. server according to claim 26, wherein, the computer program instructions also include：

28. server according to claim 27, wherein, the computer program instructions also include：

29. according to the server described in claim 26 to 28 any one, wherein, the pretreatment includes following any：

30. according to the server described in claim 25 to 28 any one, wherein, the historical trading data also includes：Hand over The easy time.

31. server according to claim 30, wherein, first predetermined number by corresponding to target transaction shop The characteristic vector of trade user include as the transaction feature vector in the target transaction shop：

32. server according to claim 30, wherein, second predetermined number by corresponding to target transaction user The characteristic vector in transaction shop include as the transaction feature vector of the target transaction user：

33. according to the server described in claim 25 to 28 any one, wherein, the computer program instructions also include：

34. server according to claim 28, wherein, the computer program instructions also include：

35. according to the server described in claim 25 to 28 any one, wherein, the computer program instructions also include：

36. server according to claim 28, wherein, the computer program instructions also include：