CN115619496A

CN115619496A - Accurate pushing method for E-commerce products and server

Info

Publication number: CN115619496A
Application number: CN202211144138.XA
Authority: CN
Inventors: 郑凯
Original assignee: Shenzhen Mobi E Commerce Co ltd
Current assignee: Shenzhen Mobi E Commerce Co ltd
Priority date: 2022-09-20
Filing date: 2022-09-20
Publication date: 2023-01-17

Abstract

According to the accurate pushing method and the server for the E-commerce product, provided by the embodiment of the invention, the interactive event knowledge field of the E-commerce activity log of the user to be analyzed is obtained by extracting the interactive event knowledge of the E-commerce activity log of the user to be analyzed; then, portrait label reasoning processing is carried out on the interactive event knowledge field of the E-commerce activity log of the user to be analyzed, and Q portrait label sets are obtained; acquiring a target portrait label set of a user E-commerce activity log to be analyzed through the Q portrait label sets; and finally, obtaining a user portrait of the user corresponding to the E-commerce activity log of the user to be analyzed based on the target portrait label set, and carrying out targeted pushing on the E-commerce product according to the user portrait. The method and the device for analyzing the electronic commerce activity log of the user can reduce the operation process of the interaction event knowledge field of the electronic commerce activity log of the user to be analyzed, and increase the obtaining efficiency of the target portrait label set of the user portrait for describing the electronic commerce activity log of the user.

Description

Accurate pushing method for E-commerce products and server

Technical Field

The application relates to the field of artificial intelligence and the Internet, in particular to an accurate pushing method and server for electronic commerce products.

Background

With the rapid development and wide popularization of the internet, the e-commerce platform has become an important shopping channel for network users. How to accurately push commodities on a platform to a user is a direction in which an e-commerce platform is dedicated to research. The user can leave traces in various E-commerce activities of the E-commerce commodity platform, and the behavior traces are analyzed, so that the behavior analysis result of the user can be drawn, namely the user portrait. The user portrait is a virtual representation of a real user, and the user characteristics are visualized by establishing a target user model on real data, so that a virtual image formed by fitting corresponding tags is abstracted. The user portrait can help the e-commerce platform to make operation decision, so that how to process the e-commerce data to obtain an accurate user portrait is a research subject of the e-commerce platform.

It should be noted that the above-mentioned contents are only for facilitating the understanding of the technical solution of the present application, and are not used as a basis for evaluating the inventive idea of the present application.

Disclosure of Invention

The invention aims to provide an accurate pushing method and a server for an e-commerce product, which are used for helping to analyze a user portrait according to the e-commerce activity of the user and intelligently and accurately push corresponding services.

In order to achieve the above object, the embodiments of the present invention are implemented as follows:

in a first aspect, an embodiment of the present invention provides an accurate pushing method for e-commerce products, which is applied to a server, and the method includes:

extracting the interactive event knowledge of the E-commerce activity log of the user to be analyzed to obtain an interactive event knowledge field of the E-commerce activity log of the user to be analyzed;

carrying out portrait label reasoning processing on the interaction event knowledge field of the E-commerce activity log of the user to be analyzed to obtain Q portrait label sets;

each portrait label set in the Q portrait label sets comprises P portrait labels, the first portrait label set in the Q portrait label sets is obtained through an interaction event knowledge field of the E-commerce activity log of the user to be analyzed, the vth portrait label set in the Q portrait label sets is obtained through the top v-1 portrait label sets, and P is greater than 1; the Q is more than 1; v < Q;

acquiring a target portrait label set of the E-commerce activity log of the user to be analyzed through the Q portrait label sets;

and obtaining a user portrait of the user corresponding to the E-commerce activity log of the user to be analyzed based on the target portrait label set, and carrying out targeted pushing on E-commerce products according to the user portrait.

As a possible implementation, the performing portrait label inference processing on the interaction event knowledge field of the user e-commerce activity log to be analyzed to obtain Q portrait label sets includes:

carrying out portrait label inference processing on an interaction event knowledge field of the E-commerce activity log of the user to be analyzed and a first example portrait label set corresponding to the E-commerce activity log of the user to be analyzed to obtain a first portrait label set;

the first exemplary portrait tab set comprises L set portrait tabs, and P portrait tabs included in the first portrait tab set are acquired simultaneously, wherein L is less than or equal to P;

performing portrait label inference processing on an interactive event knowledge field of the E-commerce activity log of the user to be analyzed, the first example portrait label set and the first portrait label set to obtain a secondary portrait label set;

the P portrait tags included in the secondary portrait tag set are acquired simultaneously.

As a possible implementation, the extracting of the knowledge of the interaction event from the e-commerce activity log of the user to be analyzed to obtain the knowledge field of the interaction event of the e-commerce activity log of the user to be analyzed includes:

based on an interaction event knowledge extraction module aiming at an interaction event in a user e-commerce activity log analysis network, extracting interaction event knowledge from the e-commerce activity log of the user to be analyzed to obtain an interaction event knowledge field of the e-commerce activity log of the user to be analyzed;

and based on a coding module in the user e-commerce activity log analysis network, extracting the event environment knowledge from the interaction event knowledge field of the user e-commerce activity log to be analyzed to obtain the interaction event knowledge field of the user e-commerce activity log to be analyzed.

As a possible implementation, the performing portrait label inference processing on the interaction event knowledge field of the to-be-analyzed user e-commerce activity log and the first example portrait label set corresponding to the to-be-analyzed user e-commerce activity log to obtain the first portrait label set includes:

performing knowledge correlation processing on the interaction event knowledge field of the user E-commerce activity log to be analyzed and the first example portrait label set to obtain an inference knowledge field corresponding to the first example portrait label set;

carrying out portrait label support degree reasoning on a reasoning knowledge field corresponding to the first example portrait label set to obtain the support degree of a P group reasoning portrait label;

the support degree of each group of inference portrait labels in the support degrees of the P groups of inference portrait labels comprises: the support degree corresponding to each inference portrait label in each group of inference portrait labels;

and determining the portrait label in each group of inferred portrait labels according to the support degree of the P groups of inferred portrait labels, and acquiring the first portrait label set.

As a possible implementation, the performing knowledge association processing on the interaction event knowledge field of the user e-commerce activity log to be analyzed and the first example sketch tag set to obtain an inference knowledge field corresponding to the first example sketch tag set includes:

performing knowledge field dimension reduction operation on the first example portrait label set based on a dimension reduction unit in a user e-commerce activity log analysis network to obtain a dimension reduction knowledge field corresponding to the first example portrait label set;

performing knowledge field shielding operation on a dimension reduction knowledge field corresponding to the first example portrait label set based on a knowledge shielding unit in the user e-commerce activity log analysis network to obtain a shielding knowledge field corresponding to the first example portrait label set;

based on a cross-correlation unit in the user e-commerce activity log analysis network, carrying out cross-correlation operation on an interaction event knowledge field of the user e-commerce activity log to be analyzed and a shielding knowledge field corresponding to the first example portrait label set to obtain a correlation knowledge field corresponding to the first example portrait label set;

and performing classification mapping operation on the associated knowledge field corresponding to the first example portrait label set based on an inference mapping unit in the user E-commerce activity log analysis network to obtain the inference knowledge field corresponding to the first example portrait label set.

As a possible implementation, said determining a portrait label in each set of inferred portrait labels based on the support of said P sets of inferred portrait labels, obtaining said first set of portrait labels, comprises:

and determining the inference portrait label with the highest support degree in each group of inference portrait labels as the portrait label corresponding to each group of inference portrait labels, and obtaining the first portrait label set.

As a possible implementation, the user e-commerce activity log analysis network is obtained by debugging a preset user e-commerce activity log analysis network;

the method also comprises the step of debugging the preset user e-commerce activity log analysis network, and comprises the following steps:

obtaining debugging samples, wherein the debugging samples comprise user e-commerce activity log debugging samples and a target portrait label set of the user e-commerce activity log debugging samples;

splitting the indication portrait label set of the user e-commerce activity log debugging sample one by one on the basis of L set portrait labels in the indication portrait label set of the user e-commerce activity log debugging sample and P as an additional number to obtain u debugging portrait label sets corresponding to the user e-commerce activity log debugging sample;

the indication portrait label set of the user e-commerce activity log debugging sample is obtained by deploying the L set portrait labels and a target portrait label set of the user e-commerce activity log debugging sample, wherein u is greater than 1;

and determining the user e-commerce activity log debugging sample as loading data of the preset user e-commerce activity log analysis network, determining the u debugging portrait label sets as the loading data of the preset user e-commerce activity log analysis network one by one, determining a target portrait label set of the user e-commerce activity log debugging sample as a standard result, and debugging the preset user e-commerce activity log analysis network to obtain the user e-commerce activity log analysis network through correction.

As a possible implementation, the user e-commerce activity log analysis network is obtained by modifying and correcting a temporary user e-commerce activity log analysis network obtained after the preset user e-commerce activity log analysis network is debugged;

the method further comprises the following steps:

acquiring an improved sample of the E-commerce activity log of the user;

determining the user e-commerce activity log improvement sample as the loaded data of the temporary user e-commerce activity log analysis network to obtain w groups of image label reasoning results; wherein each of the portrait label inference results of the w groups of portrait label inference results comprises x portrait label aggregation results, each of the portrait label aggregation results of the x portrait label aggregation results comprises P portrait label aggregation results, an f-th portrait label aggregation result of the x portrait label aggregation results is obtained through a first f-1 portrait label aggregation results, and w > 1; x is more than 1, and f is less than or equal to x;

and improving and correcting the temporary user e-commerce activity log analysis network through the w groups of image label reasoning results and a preset optimization algorithm to obtain the user e-commerce activity log analysis network.

As a possible implementation, the targeted pushing of the electronic commerce product according to the user representation includes:

acquiring E-commerce product introduction information corresponding to an E-commerce product to be pushed;

determining product basic information corresponding to the information matching items in the E-commerce product introduction information, and determining the E-commerce product introduction information except the product basic information in the E-commerce product introduction information as product pushing auxiliary information;

matching the product pushing auxiliary information with a label set of a plurality of user images, and determining target introduction information corresponding to image labels of the label set in the product pushing auxiliary information;

determining candidate user images of products corresponding to the E-commerce product introduction information from the user images according to target introduction information corresponding to the user images in the product pushing auxiliary information, and acquiring user image marks Mark-1 of the candidate user images based on the matching relation between the user images and the user image marks Mark-1;

mapping the product basic information and the user portrait Mark Mark-1 into vector information, and determining the obtained vector information as a target portrait description character corresponding to the E-commerce product introduction information;

acquiring a target portrait Mark Mark-2 corresponding to the E-commerce product introduction information based on a preset matching relationship between portrait description characters and the user portrait Mark Mark-2 and the target portrait description characters, and setting the target portrait Mark Mark-2 as a user portrait Mark of a product corresponding to the E-commerce product introduction information;

and pushing the product corresponding to the E-commerce product introduction information to the user corresponding to the user portrait mark.

In a second aspect, an embodiment of the present application provides a server, which includes a processor and a memory, where the memory stores a computer program, and when the processor runs the computer program, the server implements the method described above.

In the embodiment of the application, the interactive event knowledge field of the E-commerce activity log of the user to be analyzed is obtained by extracting the interactive event knowledge from the E-commerce activity log of the user to be analyzed; then, carrying out portrait label reasoning processing on the interaction event knowledge field of the E-commerce activity log of the user to be analyzed to obtain Q portrait label sets; acquiring a target portrait label set of a user E-commerce activity log to be analyzed through the Q portrait label sets; and finally, obtaining the user portrait of the user corresponding to the E-commerce activity log of the user to be analyzed based on the target portrait label set, and carrying out targeted pushing on the E-commerce product according to the user portrait. Each portrait label set in Q portrait label sets comprises P portrait labels, the first portrait label set in the Q portrait label sets is obtained by analyzing an interaction event knowledge field of a user e-commerce activity log, the vth portrait label set in the Q portrait label sets is obtained by the first v-1 portrait label sets, wherein P is more than 1; q is more than 1; and v is less than Q. In other words, multiple operations can be performed on the interaction event knowledge field of the e-commerce activity log of the user to be analyzed to obtain a plurality of portrait label sets; and then, acquiring a target portrait label set for describing a user portrait of the user E-commerce activity log to be analyzed through the obtained plurality of portrait label sets, wherein a portrait label set covering a plurality of portrait labels is obtained for each operation of the interaction event knowledge field of the user E-commerce activity log to be analyzed, so that the operation flow of the interaction event knowledge field of the user E-commerce activity log to be analyzed can be reduced, and the acquisition efficiency of the target portrait label set for describing the user portrait of the user E-commerce activity log is improved.

In the description that follows, additional features will be set forth, in part, in the description. These features will be in part apparent to those skilled in the art upon examination of the following and the accompanying drawings, or may be learned by production or use. The features of the present application may be realized and attained by practice or use of various aspects of the methodologies, instrumentalities and combinations particularly pointed out in the detailed examples that follow.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are required to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present application and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained from the drawings without inventive effort.

The methods, systems, and/or processes of the figures are further described in accordance with the exemplary embodiments. These exemplary embodiments will be described in detail with reference to the drawings. These exemplary embodiments are non-limiting exemplary embodiments in which reference numerals represent similar mechanisms throughout the various views of the drawings.

FIG. 1 is a schematic diagram illustrating the hardware and software components in a server according to some embodiments of the present application.

Fig. 2 is a flow chart of a method for pushing e-commerce products accurately according to some embodiments of the present application.

Fig. 3 is a schematic structural diagram of a pushing apparatus according to an embodiment of the present application.

Detailed Description

In order to better understand the technical solutions of the present application, the following detailed descriptions are provided with accompanying drawings and specific embodiments, and it should be understood that the specific features in the embodiments and examples of the present application are detailed descriptions of the technical solutions of the present application, and are not limitations of the technical solutions of the present application, and in a case of no conflict, the technical features in the embodiments and examples of the present application may be combined with each other.

In the following detailed description, numerous specific details are set forth by way of examples in order to provide a thorough understanding of the relevant guidance. It will be apparent, however, to one skilled in the art that the present application may be practiced without these specific details. In other instances, well-known methods, procedures, systems, compositions, and/or circuits have been described at a relatively high-level, without detail, in order to avoid unnecessarily obscuring aspects of the present application.

These and other features, functions, methods of execution, combination of functions and parts of related elements in the structure, and economies of manufacture may become more apparent upon consideration of the following description with reference to the accompanying drawings, all of which form a part of this application. It is to be expressly understood, however, that the drawings are for the purpose of illustration and description only and are not intended as a definition of the limits of the application. It should be understood that the drawings are not to scale. It is to be expressly understood, however, that the drawings are for the purpose of illustration and description only and are not intended as a definition of the limits of the application. It should be understood that the drawings are not to scale.

Flowcharts are used herein to illustrate the implementations performed by systems according to embodiments of the present application. It should be expressly understood that the execution of the flow diagrams may be performed out of order. Rather, these implementations may be performed in the reverse order or simultaneously. In addition, at least one other implementation may be added to the flowchart. One or more implementations may be deleted from the flowchart.

Referring to fig. 1, it is a schematic structural diagram of a server 100, where the server 100 includes a pushing device 110, a memory 120, a processor 130, and a communication unit 140. The elements of the memory 120, the processor 130, and the communication unit 140 are electrically connected to each other, directly or indirectly, to enable the transmission or interaction of data. For example, the components may be electrically connected to each other via one or more communication buses or signal lines. The push device 110 includes at least one software function module which may be stored in the memory 120 in the form of software or firmware (firmware) or solidified in an Operating System (OS) of the server 100. The processor 130 is used for executing executable modules stored in the memory 120, such as software functional modules and computer programs included in the push device 110.

The Memory 120 may be, but is not limited to, a Random Access Memory (RAM), a Read Only Memory (ROM), a Programmable Read-Only Memory (PROM), an Erasable Read-Only Memory (EPROM), an electrically Erasable Read-Only Memory (EEPROM), and the like. The memory 120 is used for storing a program, and the processor 130 executes the program after receiving the execution instruction. The communication unit 140 is used to establish a communication connection between the server 100 and a terminal device that generates the e-commerce data through a network, and to transceive the data through the network.

The processor may be an integrated circuit chip having signal processing capabilities. The Processor may be a general-purpose Processor, including a Central Processing Unit (CPU), a Network Processor (NP), and the like; but may also be a Digital Signal Processor (DSP)), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components. The various methods, steps and logic blocks disclosed in the embodiments of the present invention may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

It will be appreciated that the configuration shown in fig. 1 is merely illustrative and that server 100 may include more or fewer components than shown in fig. 1 or have a different configuration than shown in fig. 1. The components shown in fig. 1 may be implemented in hardware, software, or a combination thereof.

Referring to fig. 2, a flowchart of an accurate pushing method for e-commerce products according to some embodiments of the present application is shown, and the method is applied to the server 100 in fig. 1, and may specifically include the following steps 110 to 140. On the basis of the following steps 110 to 140, alternative embodiments will be described, which should be understood as examples and not as technical features essential for implementing the present solution.

Step 110: and extracting the interactive event knowledge of the E-commerce activity log of the user to be analyzed to obtain an interactive event knowledge field of the E-commerce activity log of the user to be analyzed.

As a possible implementation manner, the to-be-analyzed user e-commerce activity log may be a collection of collected data generated by a user performing a series of e-commerce activity events (such as commodity browsing, commodity searching, commodity collection, shopping cart dynamics, commodity purchasing, recommendation information clicking, commodity evaluation, customer service interaction, and the like) through a client application program of the e-commerce platform, and it is understood that the user e-commerce activity log may be the data of the user collected according to a fixed collection period. And the interactive event knowledge field of the user E-commerce activity log to be analyzed is obtained by extracting the interactive event knowledge from the user E-commerce activity log to be analyzed and is used for depicting the vector space expression of the characteristics contained in the interactive event of the user E-commerce activity log to be analyzed. The interactive event knowledge field of the user e-commerce activity log to be analyzed can represent event knowledge of interactive event behavior data included in a data range of an interactive event (such as commodity search and commodity evaluation) concerned in the user e-commerce activity log to be analyzed, and event environment knowledge corresponding to the interactive event, wherein the event knowledge in the user e-commerce activity log to be analyzed is used for indicating feature information of the interactive event in the knowledge data range, and the event environment knowledge is used for indicating feature information of the context of the interactive event, namely, connection or influence knowledge between the interactive event and user images of other interactive events in the user e-commerce activity log to be analyzed.

As a possible implementation manner, the process of extracting the knowledge of the interaction event from the e-commerce activity log of the user to be analyzed by the server to obtain the knowledge field of the interaction event of the e-commerce activity log of the user to be analyzed specifically includes the following steps: extracting the interactive event knowledge of the E-commerce activity log of the user to be analyzed to obtain an interactive event knowledge field of the E-commerce activity log of the user to be analyzed; and extracting the event environment knowledge from the interaction event knowledge field of the E-commerce activity log of the user to be analyzed to obtain the interaction event knowledge field of the E-commerce activity log of the user to be analyzed.

Step 120: and (3) carrying out portrait label reasoning processing on the interactive event knowledge field of the E-commerce activity log of the user to be analyzed to obtain Q portrait label sets.

Each portrait label set in Q portrait label sets comprises P portrait labels, the first portrait label set in the Q portrait label sets is obtained by analyzing an interaction event knowledge field of a user e-commerce activity log, the vth portrait label set in the Q portrait label sets is obtained by the first v-1 portrait label sets, wherein P is more than 1; q is more than 1; v < Q, it is understood that each value is an integer value.

As a possible implementation manner, the process of performing portrait label inference processing on an interaction event knowledge field of a user e-commerce activity log to be analyzed by a server to obtain Q portrait label sets may specifically include the following steps: carrying out portrait label reasoning processing on an interaction event knowledge field of a user E-commerce activity log to be analyzed and a first example portrait label set corresponding to the user E-commerce activity log to be analyzed to obtain a first portrait label set; and carrying out portrait label reasoning processing on an interaction event knowledge field, a first example portrait label set and a first portrait label set of the user E-commerce activity log to be analyzed to obtain a second portrait label set. In other words, the server may perform portrait label inference on the interaction event knowledge field, the first example portrait label set, and the first v-1 portrait label sets of the user e-commerce activity log to be analyzed to obtain the vth portrait label set until the qth portrait label set is obtained. Wherein the first exemplary portrait tab set includes L set portrait tabs, where L ≦ P. P image tags contained in the first image tag set are acquired simultaneously (parallel processing), P image tags contained in the second image tag set are acquired simultaneously, and P image tags contained in each of the Q image tag sets are acquired simultaneously. In the embodiment of the present application, the set image tag is an image tag representing the start of image tag inference processing, and the content thereof may be, for example, {00}. A first exemplary portrait label set is composed of L set portrait labels configured to represent a set at which portrait label inference processing is initiated. For example, setting the value of L to 2, then the specific content of the first exemplary portrait tag set may be {00;00}. It should be noted that, if the e-commerce activity logs of the users to be analyzed are different, the number of the obtained portrait label sets may be inconsistent in the process of portrait label inference for the interaction event knowledge fields of the e-commerce activity logs of the different users to be analyzed.

As a possible implementation mode, the server conducts portrait label inference processing on an interactive event knowledge field, a first example portrait label set and a first portrait label set of a user e-commerce activity log to be analyzed, and in the process of obtaining the second portrait label set, the server can conduct portrait label inference processing on the interactive event knowledge field of the user e-commerce activity log to be analyzed and a set obtained by deploying the first example portrait label set and the first portrait label set. For example, setting P and L equal, both to 2, then a first example portrait tab set includes two setting portrait tabs, each portrait tab set containing two portrait tabs, in other words, a portrait tab set consisting of two portrait tabs may be obtained with each operation on the interaction event knowledge field of the user e-commerce activity log to be analyzed. As an example, the server performs interaction event knowledge extraction on the e-commerce activity log of the user to be analyzed to obtain an interaction event knowledge field of the e-commerce activity log of the user to be analyzed; then, based on an interaction event knowledge field of the E-commerce activity log of the user to be analyzed and a first example portrait label set [ {00;00} to process image label inference to obtain the first image label set [ {01;02} ]; interaction event knowledge fields of the user e-commerce activity log to be analyzed and sets deployed by the first example portrait label set and the first portrait label set [ {00;00;01;02} performing portrait label inference processing to obtain a second portrait label set [ {03;04} ], until a Q portrait tag set is obtained. For example, the server obtains four portrait label sets according to the above example of the log reasoning for e-commerce activity of the user to be analyzed, where the user portrait label content indicated by the first portrait label set ([ {01, { 02} ]) is [ men; a facial cleanser ], the content of the user portrait label indicated by the secondary portrait label set ([ {03; brand ], the third set of portrait labels ([ {05, { 06} ]) indicates a user portrait label content of [ high price; promotion ], a fourth set of portrait labels ([ {07; short browsing ]. Then, target portrait labels to be analyzed of the user e-commerce activity log obtained through the four portrait label sets are ([ {01; washing the face cream; an electronic product; a brand name; high price; sales promotion; at night; short browsing ].

Step 130: and acquiring a target portrait label set for analyzing the E-commerce activity log of the user through the Q portrait label sets.

As a possible implementation manner, the target portrait label set to be analyzed for the user e-commerce activity log is a text set used for depicting the user portrait to be analyzed for the user e-commerce activity log.

Step 140: and obtaining a user portrait of a user corresponding to the E-commerce activity log of the user to be analyzed based on the target portrait label set, and carrying out targeted pushing on the E-commerce product according to the user portrait.

In the embodiment of the application, user images of target users can be obtained by combining the obtained representation contents of each image label in the target image label set, for example, for the example in step 120, the target image label set to be analyzed for the user e-commerce activity log is ([ {01, {03, [ {05 } ], [ {07; washing the face cream; an electronic product; a brand name; high price; sales promotion; at night; short browsing ], based on the above portrait tags and the content of the representations, the user portrait can be inferred as: concerns about men's facial washes and electronic products, likes to browse at night for a short time, concerns about brands, and is inclined toward high-priced product sales promotion. In addition, in actual operation, the user purchasing behavior attribute may be combined with a user image (such as sex, age, place of residence, love and marriage status, academic calendar, work area, and the like) represented by a user basic attribute to depict a complete user image.

Specifically, the process of targeted pushing of e-commerce products according to user portraits may include the following steps: the method comprises the steps of obtaining E-commerce product introduction information corresponding to an E-commerce product to be pushed, determining product basic information corresponding to an information matching item in the E-commerce product introduction information, determining the E-commerce product introduction information except the product basic information in the E-commerce product introduction information as product pushing auxiliary information, matching the product pushing auxiliary information with a label set of a plurality of user images, determining target introduction information corresponding to image labels of the label set in the product pushing auxiliary information, determining candidate user images of a product corresponding to the E-commerce product introduction information from the plurality of user images according to the target introduction information corresponding to the user images in the product pushing auxiliary information, obtaining a user image label Mark-1 of the candidate user images based on the matching relation between the user images and the user image label Mark-1, mapping the product basic information and the user image label Mark-1 as vector information, determining the obtained vector information as target description characters corresponding to the E-commerce product introduction information, obtaining preset matching relations between the product description characters and the user image label Mark-2 and obtaining corresponding target description characters of the product corresponding to the E-commerce product introduction information, and pushing the target introduction information corresponding to the user images of the user images corresponding to the user images of the E-commerce product introduction information, and obtaining the E-commerce product introduction information corresponding to the user images.

It is easy to understand that the electronic commerce product is pushed based on the comparison of a plurality of user images acquired according to the above steps 110 to 140. The electronic commerce product introduction information can be introduction information for basic attributes, functions, bright spots and the like of products, the product basic information can be introduction information for commodity categories, brands, prices and the like of the products, the process is firstly carried out label matching through product pushing auxiliary information to obtain candidate user images, preliminary screening is completed, the product basic information is combined, user image marks are obtained based on preset matching relations, and accuracy of image matching can be guaranteed.

In the embodiment of the application, the server can extract the knowledge of the interaction event from the E-commerce activity log of the user to be analyzed to obtain the knowledge field of the interaction event of the E-commerce activity log of the user to be analyzed; then, portrait label reasoning processing is carried out on the interactive event knowledge field of the E-commerce activity log of the user to be analyzed, and Q portrait label sets are obtained; acquiring a target portrait label set of a user E-commerce activity log to be analyzed through the Q portrait label sets; and finally, obtaining the user portrait of the user corresponding to the E-commerce activity log of the user to be analyzed based on the target portrait label set, and carrying out targeted pushing on the E-commerce product according to the user portrait. Each portrait label set in the Q portrait label sets comprises P portrait labels, the first portrait label set in the Q portrait label sets is obtained by analyzing an interaction event knowledge field of a user e-commerce activity log, the vth portrait label set in the Q portrait label sets is obtained by the first v-1 portrait label sets, and P is greater than 1; q is more than 1; and v is less than Q. The interaction event knowledge field of the E-commerce activity log of the user to be analyzed can be operated for multiple times to obtain a plurality of portrait label sets; and then acquiring a target portrait label set for describing the user portrait of the user E-commerce activity log to be analyzed through the obtained plurality of portrait label sets, wherein a portrait label set covering a plurality of portrait labels is obtained for each operation of the interactive event knowledge field of the user E-commerce activity log to be analyzed, so that the operation flow of the interactive event knowledge field of the user E-commerce activity log to be analyzed can be reduced, and the acquisition efficiency of the target portrait label set for describing the user portrait of the user E-commerce activity log is improved.

Based on the above accurate pushing method for the e-commerce product, the embodiment of the application further provides an implementation process of another accurate pushing method for the e-commerce product, and in the implementation process, the accurate pushing method for the e-commerce product can be implemented by adopting a user e-commerce activity log analysis network. As an implementation manner, the user e-commerce activity log analysis network may be any feasible artificial intelligence model, for example, a machine learning network or a deep learning network, and may include an Encoder module that performs interactive event knowledge extraction on the user e-commerce activity log and a Decoder module corresponding to the Encoder module, where the Decoder module is configured to perform portrait tag inference processing on an interactive event knowledge field extracted by the Encoder. Specifically, the infrastructure of the user e-commerce activity log analysis network may be a Transformer, and the Encoder module of the user e-commerce activity log analysis network may be an encoding module including an interaction event knowledge extraction module for interaction events, such as a convolutional neural network module, i encoding modules which are composed of a knowledge mask unit (mask) and an inference mapping unit (e.g., a full connection mapping FCL). Corresponding to the Encoder, the Decoder of the user e-commerce activity log analysis network can comprise a dimension reduction unit (such as Embedding), i decoders, a linear unit and a standardization unit, wherein the i decoders are composed of a knowledge shielding unit, a cross-correlation unit (such as MAN) and an inference mapping unit. The Encoder and Decoder correspond in number, and specific numerical values are not limited.

Based on the network structure of the user e-commerce activity log analysis network, the specific process of executing the accurate pushing method of the e-commerce product can refer to the following steps:

step 210: and performing interaction event knowledge extraction on the user E-commerce activity log to be analyzed based on an interaction event knowledge extraction module aiming at the interaction event in the user E-commerce activity log analysis network to obtain an interaction event knowledge field of the user E-commerce activity log to be analyzed.

Step 220: and extracting the knowledge of the event environment from the interaction event knowledge field of the user E-commerce activity log to be analyzed based on a coding module in the user E-commerce activity log analysis network to obtain the interaction event knowledge field of the user E-commerce activity log to be analyzed.

In the above steps 210 to 220, the interaction event knowledge field of the e-commerce activity log of the user to be analyzed may represent event knowledge of interaction event behavior data included in a data range of an interaction event (such as commodity search and commodity evaluation) concerned in the e-commerce activity log of the user to be analyzed, and event environment knowledge corresponding to the interaction event, where the event knowledge in the e-commerce activity log of the user to be analyzed is used to indicate feature information of the interaction event in the knowledge data range, and the event environment knowledge is used to indicate feature information of a context of the interaction event, in other words, link or influence knowledge between the interaction event and a user image of another interaction event in the e-commerce activity log of the user to be analyzed.

As a possible implementation manner, the server performs event environment knowledge extraction on an interaction event knowledge field of the user e-commerce activity log to be analyzed based on a coding module in the user e-commerce activity log analysis network, the process of obtaining the interaction event knowledge field of the user e-commerce activity log to be analyzed can be processed according to i coding modules, the event environment knowledge extraction is performed on the interaction event knowledge field of the user e-commerce activity log to be analyzed based on a first coding module in the i coding modules, and then the event environment knowledge extraction is performed on the result output by the first coding module based on a second coding module until the event environment knowledge extraction is performed on the result output by the i-1 coding module based on the i coding module, so that the interaction event knowledge field of the user e-commerce activity log to be analyzed is obtained; and in the process of processing based on each coding module in the i coding modules, processing is performed based on a knowledge masking unit and an inference mapping unit in the coding modules at one time.

Step 230: and carrying out portrait label inference processing on an interactive event knowledge field of the E-commerce activity log of the user to be analyzed and a first example portrait label set corresponding to the E-commerce activity log of the user to be analyzed to obtain a first portrait label set.

The first portrait label set comprises L set portrait labels, the first portrait label set comprises P portrait labels, the first portrait label set is obtained by analyzing an interaction event knowledge field of a user E-commerce activity log, and the P portrait labels in the first portrait label set are obtained simultaneously; wherein L is less than or equal to P. As a possible implementation manner, a specific process of performing portrait label inference processing on an interaction event knowledge field of a user e-commerce activity log to be analyzed and a first example portrait label set corresponding to the user e-commerce activity log to be analyzed by a server to obtain a first portrait label set may include the following steps: carrying out knowledge association processing on an interaction event knowledge field and a first example portrait label set of a user E-commerce activity log to be analyzed to obtain an inference knowledge field corresponding to the first example portrait label set; carrying out portrait label support degree reasoning on a reasoning knowledge field corresponding to the first example portrait label set to obtain the support degree of the P groups of reasoning portrait labels; the support degree of each group of inference portrait labels in the support degrees of the P groups of inference portrait labels comprises the support degree corresponding to each inference portrait label in each group of inference portrait labels; and determining the portrait label in each group of inferred portrait labels based on the support degree of the P groups of inferred portrait labels to obtain a first portrait label set. The method comprises the steps that an inference knowledge field corresponding to a first example portrait label set is used for indicating that the inference knowledge field at the moment is obtained from the first example portrait label, and the obtained portrait labels are all used for portraying a user E-commerce activity log to be analyzed; in addition, the method also includes obtaining a reasoning knowledge field corresponding to the first portrait label set from the first portrait label set, so as to obtain a second portrait label set. As a possible implementation manner, the specific process of the server performing knowledge association processing on the interaction event knowledge field of the user e-commerce activity log to be analyzed and the first example portrait label set to obtain the inference knowledge field corresponding to the first example portrait label set includes: performing knowledge field dimension reduction operation on the first example portrait label set based on a dimension reduction unit in a user E-commerce activity log analysis network to obtain a dimension reduction knowledge field corresponding to the first example portrait label set; and processing the interaction event knowledge field of the E-commerce activity log to be analyzed and the dimension reduction knowledge field corresponding to the first example portrait label set based on i decoders in the user E-commerce activity log analysis network to obtain an inference knowledge field corresponding to the first example portrait label set. As a possible implementation manner, the server may process the interaction event knowledge field of the e-commerce activity log of the user to be analyzed and the dimension reduction knowledge field corresponding to the first example sketch tag set based on a first Decoder of i decoders, and then process the interaction event knowledge field of the e-commerce activity log of the user to be analyzed and a result generated by the first Decoder through a second Decoder until the interaction event knowledge field of the e-commerce activity log of the user to be analyzed and the result generated by the i-1 Decoder are processed according to the i-th Decoder, so as to obtain the inference knowledge field corresponding to the first example sketch tag set; when processing is carried out according to each Decoder in the i decoders, the processing is carried out based on a knowledge shielding unit, a cross-correlation unit and an inference mapping unit in the decoders, and an interaction event knowledge field of a user E-commerce activity log is analyzed and obtained by processing according to the cross-correlation unit loaded in each Decoder. The server can perform output dimension reduction operation and position dimension reduction operation on the first example portrait label set in the process of performing knowledge field dimension reduction operation on the first example portrait label set by a dimension reduction unit in the user e-commerce activity log analysis network.

As a possible implementation manner, if the user e-commerce activity log analysis network is established by a Decoder, the server performs knowledge association processing on the interaction event knowledge field to be analyzed for the user e-commerce activity log and the first example portrait label set, and a process of obtaining the inference knowledge field corresponding to the first example portrait label set may specifically include the following steps: performing knowledge field dimension reduction operation on the first example portrait label set based on a dimension reduction unit in a user E-commerce activity log analysis network to obtain a dimension reduction knowledge field corresponding to the first example portrait label set; performing knowledge field shielding operation on a dimension reduction knowledge field corresponding to the first example portrait label set based on a knowledge shielding unit in a user E-commerce activity log analysis network to obtain a shielding knowledge field corresponding to the first example portrait label set; based on a cross-correlation unit in a user e-commerce activity log analysis network, carrying out cross-correlation operation on an interaction event knowledge field of a user e-commerce activity log to be analyzed and a shielding knowledge field corresponding to a first example portrait label set to obtain a correlation knowledge field corresponding to the first example portrait label set; and performing classification mapping operation on the associated knowledge fields corresponding to the first example portrait label set based on a reasoning mapping unit in the user E-commerce activity log analysis network to obtain reasoning knowledge fields corresponding to the first example portrait label set.

As an embodiment, the process of the server performing portrait label support inference on the inference knowledge field corresponding to the first example portrait label set to obtain the support of the P groups of inferred portrait labels may specifically include the following steps: performing linear transformation on an inference knowledge field corresponding to the first example portrait label set based on a linear unit in the user E-commerce activity log analysis network; and then, loading the transformation result into an activation function for standardization processing to obtain the support degree of the P groups of inference portrait labels. In this embodiment of the application, the support degree of each group of inference portrait tags in the support degrees of P groups of inference portrait tags includes a support degree (probability) corresponding to each inference portrait tag in each group of inference portrait tags, and each inference portrait tag in each group of inference portrait tags is selected from candidate portrait tags, for example, a portrait tag that is not 0 in the support degrees of the candidate portrait tags based on the user e-commerce activity log analysis network inference. In other words, the server maps the transformation result of the linear unit into the support degrees of P groups of candidate portrait labels based on an activation function in the user E-commerce activity log analysis network, the support degrees of all the candidate portrait labels are covered in the support degrees of each group of candidate portrait labels, then the candidate portrait labels with the support degrees of each group of candidate portrait labels different from 0 are determined as inference portrait labels, and the support degrees of the P groups of inference portrait labels are obtained.

As an implementation mode, the process that the server determines the portrait label in each group of inferred portrait labels through the support degree of the P groups of inferred portrait labels to obtain the first portrait label set specifically includes the following steps: and determining the inference portrait label with the highest support degree in each group of inference portrait labels as the portrait label corresponding to each group of inference portrait labels to obtain a first portrait label set. For example, setting P and L equal to each other and 2, that is, the first exemplary portrait tab set includes two setting portrait tabs, each portrait tab set includes two portrait tabs, and a portrait tab set composed of two portrait tabs is obtained upon each operation on the interaction event knowledge field of the user E-commerce activity log to be analyzed. As an example; the server extracts the interactive event knowledge of the user E-commerce activity log to be analyzed based on an interactive event knowledge extraction module aiming at the interactive event in the user E-commerce activity log analysis network to obtain an interactive event knowledge field of the user E-commerce activity log to be analyzed; extracting event environment knowledge from an interactive event knowledge field of a user E-commerce activity log to be analyzed based on i encoders in the user E-commerce activity log analysis network to obtain the interactive event knowledge field of the user E-commerce activity log to be analyzed; analyzing a dimension reduction unit in the network based on the user e-commerce activity log, and performing comparison on a first example portrait label set [ {00;00} ] performing a knowledge field dimension reduction operation to obtain a dimension reduction knowledge field corresponding to the first example portrait label set; then, processing an interaction event knowledge field of the user E-commerce activity log to be analyzed and a dimension reduction knowledge field corresponding to a first example portrait label set based on i decoders in the user E-commerce activity log analysis network to obtain a reasoning knowledge field corresponding to the first example portrait label set, wherein the cross association unit in each Decoder is used for completing the association interaction with the interaction event knowledge field of the user E-commerce activity log to be analyzed; then, based on a linear unit in the user E-commerce activity log analysis network, a reasoning knowledge field corresponding to the first example portrait label set is transformed, a transformation result is loaded into an activation function for standardization processing, and the support degree of two groups of reasoning portrait labels is obtained; and determining the inference portrait label with the highest support degree in each group of inference portrait labels as the portrait label corresponding to the group of inference portrait labels, thereby obtaining a first portrait label set. If the content corresponding to the support degree of the first group of inference portrait labels is [ men; boy ], the first set of inference sketch labels has a support of [0.5;0.3], the corresponding content of the support degree of the second group of inference portrait labels is [ facial cleanser; emulsion in water ], the first set of inferred portrait tags have a support of [0.7;0.6], then the image label identified in the first set of inferred image labels is [ male ], the image label identified in the second set of inferred image labels is [ facial cleanser ], and the resulting first set of image labels is [ male; facial cleanser, [ men; facial washes ] may be represented by numerical labels.

Step 240: and (3) performing portrait label reasoning processing on an interactive event knowledge field, a first example portrait label set and a first portrait label set of the user E-commerce activity log to be analyzed to obtain a secondary portrait label set until a Qth portrait label set is obtained.

The P portrait tags included in the secondary portrait tag sets are concurrently acquired, and the P portrait tags included in each portrait tag set are concurrently acquired. As a possible implementation manner, the server performs portrait label inference processing on an interaction event knowledge field, a first example portrait label set and a first portrait label set of the user e-commerce activity log to be analyzed, and the process of obtaining the secondary portrait label set may be obtained by performing portrait label inference processing on the interaction event knowledge field of the user e-commerce activity log to be analyzed and a set obtained by deploying the first example portrait label set and the first portrait label set; in other words, the deployed sets of the first example representation tag set and the first representation tag set are loaded into the Decoder of the user e-commerce activity log analysis network.

Step 250: and acquiring a target portrait label set for analyzing the E-commerce activity log of the user through the Q portrait label sets.

In the accurate pushing method of the E-commerce product provided by the embodiment of the application, the server extracts the knowledge of the interaction event from the E-commerce activity log of the user to be analyzed based on the user E-commerce activity log analysis network to obtain the knowledge field of the interaction event of the E-commerce activity log of the user to be analyzed; performing portrait label inference processing on an interaction event knowledge field of the E-commerce activity log of the user to be analyzed and a first example portrait label set corresponding to the E-commerce activity log of the user to be analyzed to obtain a first portrait label set; carrying out portrait label reasoning processing on an interaction event knowledge field, a first example portrait label set and a first portrait label set of a user E-commerce activity log to be analyzed to obtain a secondary portrait label set until a Q portrait label set is obtained; obtaining a target portrait label set of a user E-commerce activity log to be analyzed through the Q portrait label sets; wherein a first exemplary portrait tab set includes L set portrait tabs, each portrait tab set includes P portrait tabs, the P portrait tabs included in each portrait tab set are captured simultaneously, P > 1, Q > 1, L ≦ P. In other words, multiple operations can be performed on the interaction event knowledge field of the user e-commerce activity log to be analyzed based on the user e-commerce activity log analysis network to obtain a plurality of portrait label sets; and then acquiring a target portrait label set for depicting a user portrait of the user E-commerce activity log to be analyzed through the acquired plurality of portrait label sets, wherein a portrait label set covering a plurality of portrait labels is acquired by each operation on an interaction event knowledge field of the user E-commerce activity log to be analyzed, and a plurality of portrait labels in the portrait label set are acquired simultaneously, so that the operation flow of the interaction event knowledge field of the user E-commerce activity log to be analyzed can be reduced, and the acquisition efficiency of the target portrait label set for depicting the user portrait of the user E-commerce activity log is improved.

The user e-commerce activity log analysis network is obtained by debugging a preset user e-commerce activity log analysis network, the preset user e-commerce activity log analysis network and the user e-commerce activity log analysis network comprise a consistent network architecture, but the network coefficient of the preset user e-commerce activity log analysis network needs to be adjusted to meet the application requirement. The embodiment of the application also provides a debugging method for the preset user e-commerce activity log analysis network, and the debugging process can comprise the following steps:

step 310: and obtaining a debugging sample.

In an embodiment of the application, the debugging samples comprise user e-commerce activity log debugging samples and a target portrait label set of the user e-commerce activity log debugging samples. The user e-commerce activity log debugging sample is used for debugging a preset user e-commerce activity log analysis network, and is different from a user e-commerce activity log to be analyzed and points to a user e-commerce activity log of a random user portrait. The target portrait label set of the user e-commerce activity log debugging sample is a label set of a user portrait used to depict the user e-commerce activity log debugging sample.

Step 320: the method comprises the steps of taking L set portrait labels included in an indication portrait label set of a user e-commerce activity log debugging sample as a basis, taking P as an additional number, splitting the indication portrait label set of the user e-commerce activity log debugging sample one by one, and obtaining u debugging portrait label sets corresponding to the user e-commerce activity log debugging sample.

And the indication portrait label set of the user e-commerce activity log debugging sample is obtained by deploying L set portrait labels and a target portrait label set of the user e-commerce activity log debugging sample, wherein u is larger than 1. Each debugging portrait label set in u debugging portrait label sets corresponding to the user E-commerce activity log debugging samples comprises L setting portrait labels. It should be noted that, if the user e-commerce activity log debugging samples are different, the collection capacities of the indicated portrait label sets of the different user e-commerce activity log debugging samples obtained by deployment according to the target portrait label sets of the different user e-commerce activity log debugging samples may be different. If L is fixed, in the process of splitting the indication portrait label sets of the user e-commerce activity log debugging samples with different set capacities, the number of the debugging portrait label sets corresponding to the obtained user e-commerce activity log debugging samples is inconsistent.

For example, setting P and L equal to each other and both equal to 2, if the target portrait label set of the user e-commerce activity log debugging sample is [ ], the indication portrait label set of the user e-commerce activity log debugging sample is [ male; washing the face cream; at night; promotion ] the server divides the instruction portrait label set of the user E-commerce activity log debugging sample one by one based on two set portrait labels in the instruction portrait label set of the user E-commerce activity log debugging sample and by taking 2 as an additional number, and obtains 3 debugging portrait label sets corresponding to the user E-commerce activity log debugging sample: [00;00, 00;00; at night; promotion ], [ men; washing the face cream; at night; promotion ].

Step 330: the user e-commerce activity log debugging sample is determined as loading data of a preset user e-commerce activity log analysis network, u debugging portrait label sets are used as the loading data of the preset user e-commerce activity log analysis network one by one, a target portrait label set of the user e-commerce activity log debugging sample is used as a standard result, and the preset user e-commerce activity log analysis network is debugged to obtain the user e-commerce activity log analysis network through correction.

As a possible implementation manner, the server determines a user e-commerce activity log debugging sample as loaded data of a preset user e-commerce activity log analysis network, notifies that u debugging image label sets are determined as loaded data of the preset user e-commerce activity log analysis network one by one, determines a target image label set of the user e-commerce activity log debugging sample as a standard result, debugs the preset user e-commerce activity log analysis network, determines the user e-commerce activity log debugging sample as loaded data of the preset user e-commerce activity log analysis network when the user e-commerce activity log analysis network is obtained through correction, extracts interaction event knowledge from the user e-commerce activity log debugging sample according to the preset user e-commerce activity log analysis network, and obtains an interaction event knowledge field of the user e-commerce activity log debugging sample; determining the u debugging portrait label sets one by one as loading data of a preset user e-commerce activity log analysis network, and enabling each debugging portrait label set in the u debugging portrait label sets to be in knowledge correlation with interaction event knowledge fields of user e-commerce activity log debugging samples one by one to obtain inference knowledge fields corresponding to each debugging portrait label set one by one; and debugging the preset user e-commerce activity log analysis network according to the support of the P debugging inference portrait labels corresponding to each debugging portrait label set, the target portrait label set of the user e-commerce activity log debugging samples and the cost index, so as to correct and obtain the user e-commerce activity log analysis network. Wherein, the set formed by P debugging inference image labels corresponding to the debugging image label set is consistent with the rest debugging inference image labels except the debugging image label set in the next debugging image label set of the debugging image label set.

As a possible implementation manner, the server determines the user e-commerce activity log debugging sample as the loading data of the preset user e-commerce activity log analysis network, performs interaction event knowledge extraction on the user e-commerce activity log debugging sample based on the preset user e-commerce activity log analysis network, and obtains the interaction event knowledge field of the user e-commerce activity log debugging sample, which is the same as the above step of performing interaction event knowledge extraction on the user e-commerce activity log to be analyzed based on the user e-commerce activity log analysis network to obtain the interaction event knowledge field of the user e-commerce activity log to be analyzed.

As an implementation mode, a first debugging portrait label set in u debugging portrait label sets is determined as loading data of a preset user e-commerce activity log analysis network, and knowledge association processing is carried out on interaction event knowledge fields of the first debugging portrait label set and a user e-commerce activity log debugging sample so as to obtain inference knowledge fields corresponding to the first debugging portrait label set. As a possible implementation manner, the server may perform knowledge field dimension reduction operation on a first debugging representation tag set based on a dimension reduction unit in a preset user e-commerce activity log analysis network to obtain a dimension reduction knowledge field corresponding to the first debugging representation tag set, and process an interaction event knowledge field of a user e-commerce activity log debugging sample and a dimension reduction knowledge field corresponding to the first debugging representation tag set based on i decoders in the preset user e-commerce activity log analysis network to obtain an inference knowledge field corresponding to the first debugging representation tag set. As a possible implementation mode, the server processes an interaction event knowledge field of a user E-commerce activity log debugging sample and a dimensionality reduction knowledge field corresponding to a first debugging label set based on a first Decoder in i decoders in a preset user E-commerce activity log analysis network, and then processes the interaction event knowledge field of the user E-commerce activity log debugging sample and a result generated by the first Decoder through a second Decoder until the interaction event knowledge field of the user E-commerce activity log debugging sample and the result generated by the L-1 Decoder are processed according to the i Decoder, so that an inference knowledge field corresponding to the first debugging image label set is obtained.

As an implementation mode, when the server analyzes each Decoder in i decoders of the network based on the preset user e-commerce activity log, the server processes the i decoders sequentially based on a knowledge masking unit, a cross-correlation unit and an inference mapping unit in the Decoder. And the interaction event knowledge field of the user E-commerce activity log debugging sample is obtained by processing based on the cross-correlation unit loaded into each Decoder. As an embodiment, if the knowledge masking unit in the Decoder in the preset user e-commerce activity log analysis network is an unrestricted knowledge masking unit, when each Decoder in the i decoders of the preset user e-commerce activity log analysis network is processed, the processing may be performed sequentially based on the unrestricted knowledge masking unit, the cross-correlation unit and the inference mapping unit in the Decoder. The preset user E-commerce activity log analysis network forms a Decoder through an unlimited knowledge masking unit, an indication image label set of a user E-commerce activity log debugging sample can be Sequentially Masked (SM) based on the unlimited knowledge masking unit, the preset user E-commerce activity log analysis network can be enabled to process a debugging image label set of u debugging image label sets only based on the indication image label set of the user E-commerce activity log debugging sample, the knowledge of the debugging image label set is debugged, and the knowledge of debugging image labels outside the debugging image label set in the indication image label set of the user E-commerce activity log debugging sample is not needed.

In the process of reasoning the support degrees of the image labels one by one for the reasoning knowledge fields corresponding to each debugging image label set and obtaining the support degrees of P debugging inference image labels corresponding to each debugging image label set, the debugging image label set is identical to the debugging image label set except for the debugging image label set in the set formed by the P debugging inference image labels corresponding to the debugging image label set and the next debugging image label set of the debugging image label set. As a possible implementation manner, taking the first debugging portrait label set as an example, the server maps the linear transformation result to the support degrees of P groups of alternative portrait labels based on an activation function in a preset user e-commerce activity log analysis network, where the support degrees of each group of alternative portrait labels include the support degree corresponding to each alternative portrait label in all the alternative portrait labels, so as to obtain the debugging inference portrait label corresponding to each group of alternative portrait labels in the support degrees of each group of alternative portrait labels, and obtain the support degrees of P debugging inference portrait labels corresponding to the first debugging portrait label set.

As an implementation mode, the server debugs the preset user e-commerce activity log analysis network based on the support degrees of P debugging inference portrait tags corresponding to each debugging portrait tag set, the target portrait tag set of a user e-commerce activity log debugging sample and the cost index, in the process of obtaining the user e-commerce activity log analysis network through correction, the cost value of the cost index can be determined through the support degrees of the P debugging inference portrait tags corresponding to each debugging portrait tag set and the target portrait tag set of the user e-commerce activity log debugging sample, and the preset user e-commerce activity log analysis network is debugged to obtain the user e-commerce activity log analysis network through correction. The specific form and selection of the cost function may be configured and selected according to actual needs, which is not limited in the embodiment of the present application.

As a possible implementation manner, the user e-commerce activity log analysis network may be obtained by debugging a preset user e-commerce activity log analysis network, or when the preset user e-commerce activity log analysis network is debugged to obtain a temporary user e-commerce activity log analysis network, the temporary user e-commerce activity log analysis network is obtained by modifying and correcting the temporary user e-commerce activity log analysis network, and the temporary user e-commerce activity log analysis network and the user e-commerce activity log analysis network have a network architecture in accordance with the data of europe, and have a network coefficient that is inconsistent. The server improves and corrects the temporary user e-commerce activity log analysis network, and the step of obtaining the user e-commerce activity log analysis network may include the following steps: acquiring an improved sample of the E-commerce activity log of the user; determining the user e-commerce activity log improvement sample as loading data of a temporary user e-commerce activity log analysis network to obtain w groups of pictorial label reasoning results; wherein, each portrait label inference result in the w groups of portrait label inference results comprises x portrait label set results, each portrait label set result in the x portrait label set results comprises P portrait label results, the f th portrait label set result in the x portrait label set results is obtained by the f-1 portrait label set results, and w is larger than 1; x is more than 1, f is less than or equal to x. And improving and correcting the temporary user e-commerce activity log analysis network through the w groups of image label reasoning results and a preset optimization algorithm to obtain the user e-commerce activity log analysis network. In the embodiment of the application, the user e-commerce activity log improvement sample can be a user e-commerce activity log which is configured to improve and correct a temporary user e-commerce activity log analysis network, is inconsistent with the user e-commerce activity log debugging sample and the user e-commerce activity log to be analyzed and has random user portrait.

As a possible implementation manner, the server determines the user e-commerce activity log improvement sample as loaded data of a temporary user e-commerce activity log analysis network, each group of portrait label inference results in the obtained w groups of portrait label inference results are portrait labels used for describing user portrait of the user e-commerce activity log improvement sample, which are obtained based on inference of the temporary user e-commerce activity log analysis network, one group of portrait label inference results are a portrait label set, and in the obtained w groups of portrait label inference results, each group of portrait label inference results are inconsistent. It should be noted that, if the user e-commerce activity log improvement samples are not consistent, different user e-commerce activity log improvement samples are determined as the loaded data of the temporary user e-commerce activity log analysis network, and in w groups of image tag inference results corresponding to the different user e-commerce activity log improvement samples, the number of image tag set results covered by each group of image tag inference results may not be consistent. Different improvement correction requirements of the network can be analyzed through the temporary user e-commerce activity log, and a specific numerical value corresponding to w is configured. As a possible implementation manner, the server may load the user e-commerce activity log improvement sample w times in the temporary user e-commerce activity log analysis network, so that the temporary user e-commerce activity log analysis network is used to process the user e-commerce activity log improvement sample w times to obtain w groups of pictorial label inference results.

In the method, the process of processing the improved sample of the user e-commerce activity log to obtain a group of portrait label reasoning results at one time based on the temporary user e-commerce activity log analysis network is similar to the process of processing the user e-commerce activity log to be analyzed to obtain a target portrait label set of the user e-commerce activity log to be analyzed based on the user e-commerce activity log analysis network. The difference lies in that if any portrait label set in Q portrait label sets is obtained by processing the user E-commerce activity log to be analyzed based on a user E-commerce activity log analysis network, the server determines the inference portrait label with the highest support degree in each group of inference portrait labels in P groups of inference portrait labels corresponding to the portrait label set as the portrait label corresponding to each group of inference portrait labels, so as to obtain the portrait label set. When w times of processing are carried out on the user E-commerce activity log improved sample based on the temporary user E-commerce activity log analysis network to obtain w groups of portrait label reasoning results, a random portrait label of the w groups of portrait label reasoning results is obtained through a preset strategy according to the support degree of a group of result reasoning portrait labels corresponding to the random portrait label. The set of result inference portrait labels are portrait labels with support degree of not 0 in a plurality of candidate portrait labels obtained by inference on the user E-commerce activity log improvement samples based on the temporary user E-commerce activity log analysis network. The preset strategy may be self-setting, such as randomly acquiring, or sorting and then selecting a predetermined order.

In the accurate pushing method of the E-commerce product provided by the embodiment of the application, the server debugs the preset user E-commerce activity log analysis network through the debugging samples of the target portrait label set which covers the user E-commerce activity log debugging samples and the user E-commerce activity log debugging samples, and the user E-commerce activity log analysis network is obtained. The method comprises the steps of dividing an indication portrait label set of a user e-commerce activity log debugging sample one by one on the basis of L set portrait labels covered in the indication portrait label set of the user e-commerce activity log debugging sample, taking P as an additional number, obtaining u debugging portrait label sets corresponding to the user e-commerce activity log debugging sample, determining the user e-commerce activity log debugging sample as loading data of a preset user e-commerce activity log analysis network, taking the u debugging portrait label sets as loading data of the preset user e-commerce activity log analysis network one by one, determining a target portrait label set of the user e-commerce activity log debugging sample as a standard result, debugging the preset user e-commerce activity log analysis network, and correcting to obtain the user e-commerce activity log analysis network. Each debugging portrait label set in the u debugging portrait label sets comprises P debugging portrait labels, and P is more than 1; u is more than 1; l is less than or equal to P. The method has the advantages that Q portrait label sets corresponding to the E-commerce activity log of the user to be analyzed can be obtained through reasoning according to an interaction event knowledge field of the E-commerce activity log of the user to be analyzed and L set portrait label portrait labels in the process of processing the E-commerce activity log of the user to be analyzed by the user E-commerce activity log analysis network obtained based on debugging, each portrait label set in the Q portrait label sets comprises P portrait labels, and the operation process (times) of the user E-commerce activity log analysis network on the interaction event knowledge field of the E-commerce activity log of the user to be analyzed is effectively reduced, so that the speed of the user E-commerce activity log analysis network obtaining a target portrait label set of a user portrait for describing the E-commerce activity log of the user is increased, and the efficiency is improved. In addition, the temporary user e-commerce activity log analysis network obtained after the preset user e-commerce activity log analysis network is debugged is improved and corrected through a preset optimization algorithm to obtain the user e-commerce activity log analysis network, and the user e-commerce activity log analysis network obtained through the method is more accurate in obtaining a target portrait label set of a user portrait used for describing the user e-commerce activity log.

Referring to fig. 3, which is a schematic structural diagram of a pushing device 110 according to an embodiment of the present invention, the pushing device 110 may be used for executing an accurate pushing method of an e-commerce product, wherein the pushing device 110 includes:

the knowledge extraction module 111 is configured to extract the knowledge of the interaction event from the e-commerce activity log of the user to be analyzed, and obtain an interaction event knowledge field of the e-commerce activity log of the user to be analyzed.

And the label reasoning module 112 is used for carrying out portrait label reasoning processing on the interactive event knowledge field of the E-commerce activity log of the user to be analyzed to obtain Q portrait label sets.

Each portrait label set in the Q portrait label sets comprises P portrait labels, the first portrait label set in the Q portrait label sets is obtained through an interaction event knowledge field of the E-commerce activity log of the user to be analyzed, the vth portrait label set in the Q portrait label sets is obtained through the top v-1 portrait label sets, and P is greater than 1; the Q is more than 1; and v is less than Q.

And the label determining module 113 is configured to obtain a target portrait label set of the e-commerce activity log of the user to be analyzed through the Q portrait label sets.

And the pushing module 114 is used for obtaining the user portrait of the user corresponding to the e-commerce activity log of the user to be analyzed based on the target portrait label set, and performing targeted pushing on e-commerce products according to the user portrait.

The knowledge extraction module 111 may be configured to perform step 110, the tag inference module 112 may be configured to perform step 120, the tag determination module 113 may be configured to perform step 130, and the push module 114 may be configured to perform step 140.

Since the above embodiment has described the precise pushing method for the e-commerce product provided by the embodiment of the present invention in detail, and the principle of the pushing device 110 is the same as that of the method, the implementation principle of each module of the pushing device 110 is not described herein again.

In the embodiments provided in the present application, it should be understood that the disclosed apparatus and method can be implemented in other ways. The apparatus embodiments described above are merely illustrative and, for example, the flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

In addition, the functional modules in the embodiments of the present invention may be integrated together to form an independent part, or each module may exist alone, or two or more modules may be integrated to form an independent part.

The functions, if implemented in the form of software functional modules and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk, and various media capable of storing program codes.

It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrases "comprising a," "8230," "8230," or "comprising" does not exclude the presence of additional like elements in a process, method, article, or apparatus that comprises the element.

It should be understood that, for technical terms which are not noun explanations above, those skilled in the art can make a deduction from the above disclosure to determine the meaning of the named generation, such as for some values, coefficients, weights, indexes, factors and other terms, those skilled in the art can make a deduction and determination according to the logical relationship between the above and the following, and the value range of these values can be selected according to the actual situation, and is not limited herein.

The skilled person can unambiguously determine some preset, reference, predetermined, set and target technical features/terms, such as threshold values, threshold intervals, threshold ranges, etc., from the above disclosure. For some technical characteristic terms which are not explained, the technical solution can be clearly and completely implemented by those skilled in the art by reasonably and unambiguously deriving the technical solution based on the logical relations in the previous and following paragraphs. The above disclosure of the embodiments of the present application will be apparent to those skilled in the art from the above disclosure. It should be understood that the process of deriving and analyzing technical terms, which are not explained, by those skilled in the art based on the above disclosure is based on the contents described in the present application, and thus the above contents are not an inventive judgment of the overall scheme.

Having thus described the basic concept, it will be apparent to those skilled in the art that the foregoing detailed disclosure is to be considered merely illustrative and not restrictive of the broad application. Various modifications, improvements and adaptations to the present application may occur to those skilled in the art, although not explicitly described herein. Such modifications, improvements and adaptations are proposed in the present application and thus fall within the spirit and scope of the exemplary embodiments of the present application.

Also, this application uses specific terminology to describe embodiments of the application. Reference throughout this specification to "one embodiment," "an embodiment," and/or "some embodiments" means that a particular feature, structure, or characteristic described in connection with at least one embodiment of the present application is included in at least one embodiment of the present application. Therefore, it is emphasized and should be appreciated that two or more references to "an embodiment" or "one embodiment" or "an alternative embodiment" in various portions of this specification are not necessarily to the same embodiment. Furthermore, some features, structures, or characteristics of at least one embodiment of the present application may be combined as appropriate.

In addition, those skilled in the art will recognize that the various aspects of the present application may be illustrated and described in terms of any number of patentable categories or situations, including any new and useful combination of procedures, machines, products, or materials, or any new and useful modifications thereof. Accordingly, various aspects of the present application may be embodied entirely in hardware, entirely in software (including firmware, resident software, micro-code, etc.) or in a combination of hardware and software. The above hardware or software may be referred to as a "unit", "component", or "system". Furthermore, aspects of the present application may be represented as a computer product, including computer readable program code, embodied in at least one computer readable medium.

A computer readable signal medium may comprise a propagated data signal with computer program code embodied therein, for example, on a baseband or as part of a carrier wave. The propagated signal may take any of a variety of forms, including electromagnetic, optical, and the like, or any suitable combination. A computer readable signal medium may be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code on a computer readable signal medium may be propagated over any suitable medium, including radio, electrical cable, fiber optic cable, RF, or the like, or any combination of the preceding.

Computer program code required for the execution of aspects of the present application may be written in any combination of one or more programming languages, including object oriented programming, such as Java, scala, smalltalk, eiffel, JADE, emerald, C + +, C #, VB.NET, python, and the like, or similar conventional programming languages, such as the "C" programming language, visual Basic, fortran 2003, perl, COBOL 2002, PHP, ABAP, dynamic programming languages, such as Python, ruby, and Groovy, or other programming languages. The programming code may execute entirely on the user's computer, as a stand-alone software package, partly on the user's computer, partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer may be connected to the user's computer through any form of network, such as a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet), or in a cloud computing environment, or as a service using, for example, software as a service (SaaS).

Additionally, the order of the process elements and sequences described herein, the use of numerical letters, or other designations are not intended to limit the order of the processes and methods unless otherwise indicated in the claims. While various presently believed to be useful embodiments of the invention have been discussed in the foregoing disclosure by way of illustration, it is to be understood that such detail is solely for that purpose and that the appended claims are not limited to the disclosed embodiments, but, on the contrary, are intended to cover all modifications and equivalent arrangements that are within the spirit and scope of the embodiments of the disclosure. For example, although the system components described above may be implemented by hardware means, they may also be implemented by software-only solutions, such as installing the described system on an existing server or mobile device.

It should also be appreciated that in the foregoing description of embodiments of the present application, various features are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure aiding in the understanding of at least one embodiment of the invention. However, this method of disclosure is not intended to require more features than are expressly recited in the claims. Indeed, the embodiments may be characterized as having less than all of the features of a single disclosed embodiment.

Claims

1. The accurate pushing method for the E-commerce product is applied to a server and comprises the following steps:

carrying out portrait label reasoning processing on the interactive event knowledge field of the E-commerce activity log of the user to be analyzed to obtain Q portrait label sets;

each portrait label set in the Q portrait label sets comprises P portrait labels, the first portrait label set in the Q portrait label sets is obtained through an interaction event knowledge field of the E-commerce activity log of the user to be analyzed, the vth portrait label set in the Q portrait label sets is obtained through the top v-1 portrait label sets, and P is greater than 1; q is more than 1; v < Q;

2. The method of claim 1, wherein performing portrait label inference on the interaction event knowledge field of the user e-commerce activity log to be analyzed to obtain Q sets of portrait labels comprises:

carrying out portrait label reasoning processing on the interaction event knowledge field of the E-commerce activity log of the user to be analyzed, the first example portrait label set and the first portrait label set to obtain a second portrait label set;

3. The method of claim 1, wherein the extracting of the knowledge of the interaction event from the user e-commerce activity log to be analyzed to obtain the knowledge field of the interaction event of the user e-commerce activity log to be analyzed comprises:

4. The method of claim 2, wherein the portrait label inference processing is performed on the interaction event knowledge field of the user e-commerce activity log to be analyzed and a first example portrait label set corresponding to the user e-commerce activity log to be analyzed to obtain the first portrait label set, and the portrait label set includes:

carrying out portrait label support degree reasoning on a reasoning knowledge field corresponding to the first example portrait label set to obtain support degree of P groups of reasoning portrait labels;

5. The method of claim 4, wherein the performing knowledge correlation processing on the interaction event knowledge field of the user e-commerce activity log to be analyzed and the first example portrait label set to obtain an inference knowledge field corresponding to the first example portrait label set comprises:

performing knowledge field masking operation on the dimension reduction knowledge field corresponding to the first example portrait label set based on a knowledge masking unit in the user e-commerce activity log analysis network to obtain a masked knowledge field corresponding to the first example portrait label set;

and performing classification mapping operation on the associated knowledge field corresponding to the first example portrait label set based on an inference mapping unit in the user E-commerce activity log analysis network to obtain an inference knowledge field corresponding to the first example portrait label set.

6. The method of claim 4, wherein determining a portrait label in each set of inferred portrait labels in accordance with a degree of support of the P sets of inferred portrait labels, obtaining the first set of portrait labels, comprises:

7. The method of claim 1, wherein the user e-commerce activity log analysis network is obtained by debugging a preset user e-commerce activity log analysis network;

8. The method according to claim 7, wherein the user e-commerce activity log analysis network is obtained by modifying and correcting a temporary user e-commerce activity log analysis network obtained after debugging the preset user e-commerce activity log analysis network;

the method further comprises the following steps:

acquiring an improved sample of the E-commerce activity log of the user;

determining the user e-commerce activity log improvement sample as the loaded data of the temporary user e-commerce activity log analysis network to obtain w groups of image label reasoning results; wherein each portrait label inference result of the w groups of portrait label inference results comprises x portrait label set results, each portrait label set result of the x portrait label set results comprises P portrait label results, the f th portrait label set result of the x portrait label set results is obtained through the first f-1 portrait label set results, and w > 1; x is more than 1, and f is less than or equal to x;

9. The method of claim 1, wherein said pushing an e-commerce product specifically against said user representation comprises:

according to target introduction information corresponding to the user portrait in the product pushing auxiliary information, determining a candidate user portrait of a product corresponding to the E-commerce product introduction information from the plurality of user portraits, and acquiring a user portrait Mark Mark-1 of the candidate user portrait based on a matching relation between the user portrait and a user portrait Mark Mark-1;

acquiring a target user image Mark Mark-2 corresponding to the E-commerce product introduction information based on a preset matching relationship between the image description characters and the user image Mark Mark-2 and the target image description characters, and setting the target user image Mark Mark-2 as a user image Mark of a product corresponding to the E-commerce product introduction information;

10. A server, characterized by comprising a processor and a memory, said memory storing a computer program which, when executed by said processor, carries out the method according to any one of claims 1 to 9.