CN111028065A

CN111028065A - Information pushing method and device, storage medium and equipment

Info

Publication number: CN111028065A
Application number: CN201911304690.9A
Authority: CN
Inventors: 韩东亮; 徐诚浪; 莫国龙
Original assignee: Beijing Miss Fresh Electronic Commerce Co Ltd
Current assignee: Beijing Miss Fresh Electronic Commerce Co Ltd
Priority date: 2019-12-17
Filing date: 2019-12-17
Publication date: 2020-04-17

Abstract

The application discloses an information pushing method, an information pushing device, a storage medium and equipment, and belongs to the technical field of internet. The method comprises the following steps: determining a target preposed bin corresponding to a user, wherein the target preposed bin is a preposed bin in which the user has a product transaction behavior; acquiring M types of user preference products of a target front-end bin, wherein the M types of user preference products are determined according to the user preference products of each user under the target front-end bin; acquiring N hot sale products of the target preposed bin, wherein the N hot sale products are determined according to the product transaction amount under the target preposed bin; and determining products to be pushed for the user according to the M types of user preference products and the N types of hot sales products, and generating and sending pushing information of the products to be pushed to the user. The information pushing method and the information pushing device have the advantages that the information pushing is achieved based on the user preference products and the hot sales products of the front bin associated with the user, the problem that the product recommendation cannot be effectively carried out is solved, the pertinence of the information pushed to the user is strong, and the effect is good.

Description

Information pushing method and device, storage medium and equipment

Technical Field

The present application relates to the field of internet technologies, and in particular, to an information pushing method, an information pushing apparatus, a storage medium, and a device.

Background

With the vigorous development of the e-commerce industry, the fresh e-commerce platform surrounding common people dining tables is emerged at present. The fresh food e-commerce platform can supply full-grade fresh food products such as fruits and vegetables, seafood and meat, milk snacks and the like for users, and can timely distribute products purchased by the users to the hands of the users.

At present, for a fresh electric commerce platform, a problem to be solved urgently is how to recommend products to users, so that the problems that the users are buried in massive product information, and further the users run off and the like are avoided. In the related art, a collaborative filtering recommendation method or a content-based recommendation method is usually adopted to determine recommended products for a user and push corresponding information to the user. The method aims at a collaborative filtering recommendation mode, namely, users with similar preferences are utilized to recommend products, in other words, products are recommended for the users based on other users with similar preferences. The content-based recommendation method is used for recommending similar products for users based on historical preferences of the users.

In any of the above modes, the product recommendation cannot be effectively performed, and the pushed information is weak in pertinence, and the effect is poor.

Disclosure of Invention

The embodiment of the application provides an information pushing method, an information pushing device, a storage medium and equipment, which can effectively recommend products, and enable the information pushed to a user to be strong in pertinence and good in effect. The technical scheme is as follows:

in one aspect, an information pushing method is provided, where the method includes:

determining a target preposed bin corresponding to a user, wherein the target preposed bin is a preposed bin in which the user has a product transaction behavior;

acquiring M user preference products of the target front bin, wherein the M user preference products are determined according to the user preference products of each user under the target front bin, and M is a positive integer;

acquiring N hot sales products of the target preposed bin, wherein the N hot sales products are determined according to the product transaction amount under the target preposed bin, and N is a positive integer;

and determining products to be pushed for the user according to the M types of user preference products and the N types of hot sales products, generating pushing information of the products to be pushed, and sending the pushing information to the user.

In one possible implementation manner, the obtaining M user-preferred products of the target front bin includes:

determining each user with product transaction behavior in the target preposed bin according to the identification information of the target preposed bin;

obtaining user preference products of all users under the target front bin, and obtaining user preference products matched with the target front bin;

and acquiring M user preference products of the target preposed bin according to the user preference product matched with the target preposed bin.

In one possible implementation manner, the obtaining user preference products of the users under the target front bin includes:

acquiring user basic data and user behavior data of each user under the target preposed bin;

preprocessing the acquired user basic data and the user behavior data;

performing feature extraction on the preprocessed user data, inputting the generated feature data into a user preference model, and aggregating the output results of the user preference model to obtain user preference products of each user under the target preposed bin;

wherein the output result gives a preference weight value of each user for the corresponding product category.

In a possible implementation manner, the obtaining M user preference products of the target front bin according to the user preference product matched with the target front bin includes:

accumulating the quantity of the same products preferred by different users;

sorting the user preference products matched with the target front bin according to the sequence of the accumulated product quantity from large to small;

and taking the user preference products ranked at the top M as the M user preference products of the target front bin.

accumulating the same products preferred by different users according to the preference weight values;

sorting the user preference products matched with the target preposed bin according to the sequence of the accumulated preference weight values from large to small;

In one possible implementation, the method further includes:

and after the M types of user preference products are obtained, assigning new preference weight values to the M types of user preference products according to the sequence.

In one possible implementation, the method further includes:

determining a first class of users and a second class of users according to user portrait data, wherein the first class of users and the second class of users form the target user set to be subjected to information pushing;

the determining of the target front bin corresponding to the user includes: determining a target front bin corresponding to each user in the target user set;

wherein the first class of users refers to new users, and the second class of users refers to old users meeting specified conditions.

In a possible implementation manner, when the user is the first type of user, the determining, according to the M types of user preference products and the N types of hot-selling products, a product to be pushed for the user includes:

obtaining a final preference score for each of the M user preferred products and the N hot-sell products; according to the obtained final preference score, sequencing the M user preference products and the N hot sales products in a descending order;

and taking the products ranked at the top X as the products to be pushed for the user, wherein X is a positive integer.

In a possible implementation manner, when the user is the second type of user, the determining, according to the M types of user preference products and the N types of hot-selling products, a product to be pushed for the user includes:

determining a first class of products and a second class of products corresponding to the user, wherein the first class of products are historical push products, and the second class of products are products which are contrary to the user preference of the user;

filtering the first type of products and the second type of products in the M types of user preference products and the N types of hot sales products to obtain target products;

acquiring a final preference score of each product in the target products;

and sequencing the target products according to the obtained final preference score in a descending order, wherein the products sequenced at the front X positions serve as products to be pushed for the user, and X is a positive integer.

In one possible implementation, the obtaining of the final preference score includes:

acquiring preference weight values of the M types of user preference products;

acquiring hot selling weight values of the N hot selling products;

acquiring a first coefficient of the user preference product and a second coefficient of the hot-sell product;

determining the final preference score based on the obtained preference weight value and the obtained hot-sell weight value, and the obtained first coefficient and second coefficient.

In another aspect, an information pushing apparatus is provided, the apparatus including:

the first determination module is configured to determine a target front bin corresponding to a user, wherein the target front bin is a front bin in which a product transaction behavior exists for the user;

a first obtaining module configured to obtain M user preference products of the target pre-bin, where the M user preference products are determined according to user preference products of users under the target pre-bin, and M is a positive integer;

a second obtaining module configured to obtain N kinds of hot-sell products of the target pre-bin, where the N kinds of hot-sell products are determined according to a product transaction amount under the target pre-bin, and N is a positive integer;

a second determination module configured to determine products to be pushed for the user according to the M user preference products and the N hot-sell products;

the pushing module is configured to generate pushing information of the product to be pushed and send the pushing information to the user.

In a possible implementation manner, the first obtaining module is further configured to determine, according to the identification information of the target pre-bin, that each user of the product trading behavior exists in the target pre-bin; obtaining user preference products of all users under the target front bin, and obtaining user preference products matched with the target front bin; and acquiring M user preference products of the target preposed bin according to the user preference product matched with the target preposed bin.

In a possible implementation manner, the first obtaining module is further configured to obtain user basic data and user behavior data of each user under the target pre-bin; preprocessing the acquired user basic data and the user behavior data; performing feature extraction on the preprocessed user data, inputting the generated feature data into a user preference model, and aggregating the output results of the user preference model to obtain user preference products of each user under the target preposed bin;

wherein the output result gives a weight value reflecting the preference of each user for the corresponding product category.

In a possible implementation manner, the first obtaining module is further configured to accumulate product quantities of the same products preferred by different users; sorting the user preference products matched with the target front bin according to the sequence of the accumulated product quantity from large to small; and taking the user preference products ranked at the top M as the M user preference products of the target front bin.

In a possible implementation manner, the first obtaining module is further configured to accumulate the same products preferred by different users according to the preference weight values; sorting the user preference products matched with the target preposed bin according to the sequence of the accumulated preference weight values from large to small; and taking the user preference products ranked at the top M as the M user preference products of the target front bin.

In a possible implementation manner, the first obtaining module is further configured to, after obtaining the M user preference products, assign new preference weight values to the M user preference products according to a ranking.

In one possible implementation, the apparatus further includes:

a third determining module configured to determine a first class of users and a second class of users according to user portrait data, where the first class of users and the second class of users form the target user set to be subjected to information push;

the first determining module is further configured to determine a target pre-bin corresponding to each user in the target user set;

In one possible implementation, when the user is the first type of user, the second determining module is further configured to obtain a final preference score of each of the M user-preferred products and the N hot-sell products; according to the obtained final preference score, sequencing the M user preference products and the N hot sales products in a descending order; and taking the products ranked at the top X as the products to be pushed for the user, wherein X is a positive integer.

In a possible implementation manner, when the user is the second type of user, the second determining module is configured to determine a first type of product and a second type of product corresponding to the user, where the first type of product is a history push product, and the second type of product is a product that is contrary to the user preference of the user; filtering the first type of products and the second type of products in the M types of user preference products and the N types of hot sales products to obtain target products; acquiring a final preference score of each product in the target products; and sequencing the target products according to the obtained final preference score in a descending order, wherein the products sequenced at the front X positions serve as products to be pushed for the user, and X is a positive integer.

In a possible implementation manner, the second determining module is further configured to obtain preference weight values of the M user preference products; acquiring hot selling weight values of the N hot selling products; acquiring a first coefficient of the user preference product and a second coefficient of the hot-sell product; determining the final preference score based on the obtained preference weight value and the obtained hot-sell weight value, and the obtained first coefficient and second coefficient.

In another aspect, a computer-readable storage medium is provided, where at least one instruction is stored, and the at least one instruction is loaded and executed by a processor to implement the information pushing method described above.

In another aspect, an information pushing apparatus is provided, where the apparatus includes a processor and a memory, where the memory stores at least one instruction, and the at least one instruction is loaded and executed by the processor to implement the information pushing method described above.

The technical scheme provided by the embodiment of the application has the following beneficial effects:

in the information pushing process, firstly, a target preposed bin corresponding to a user is determined, wherein the target preposed bin is a preposed bin in which the user has product transaction behaviors; then, obtaining M user preference products and N hot sale products of the target preposed bin, wherein the M user preference products are determined according to the user preference products of each user under the target preposed bin, the N hot sale products are determined according to the product transaction amount under the target preposed bin, and M and N are positive integers; and then, determining products to be pushed for the user according to the M types of user preference products and the N types of hot sales products, generating pushing information of the products to be pushed, and sending the generated pushing information to the user.

Based on the above description, the information push is realized based on the user preference product and the hot sales product of the front bin associated with the user, and the method solves the problem that the product recommendation cannot be effectively carried out, so that the information push to the user is strong in pertinence and good in effect.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

Fig. 1 is a schematic diagram of a front-end bin mode provided by an embodiment of the present application;

fig. 2 is a schematic diagram of an implementation environment related to an information pushing method provided by an embodiment of the present application;

fig. 3 is a flowchart of an information pushing method provided in an embodiment of the present application;

fig. 4 is a flowchart of another information pushing method provided in an embodiment of the present application;

FIG. 5 is a flowchart for acquiring a user preference product of a user according to an embodiment of the present application;

FIG. 6 is a flow chart of obtaining user preferred products of a front bin according to an embodiment of the present application;

FIG. 7 is a flow chart of another method for obtaining user-preferred products for a pre-bin according to an embodiment of the present application;

fig. 8 is a flowchart illustrating an overall execution of an information pushing method according to an embodiment of the present application;

fig. 9 is a schematic structural diagram of an information pushing apparatus according to an embodiment of the present application;

fig. 10 is a schematic structural diagram of an information pushing apparatus according to an embodiment of the present application.

Detailed Description

To make the objects, technical solutions and advantages of the present application more clear, embodiments of the present application will be described in further detail below with reference to the accompanying drawings.

Before explaining embodiments of the present application in detail, some noun terms or abbreviations that may be referred to by the embodiments of the present application are introduced.

A front bin: also known as a micro-bin. The front-end warehouse mode is a warehouse distribution mode, each store of the fresh electric commercial platform can be a small and medium-sized storage distribution center, and therefore the central large warehouse (also called a main warehouse) only needs to supply goods to each store. After a user (also referred to as a consumer) places an order, the relevant product purchased by the user may be shipped from a nearby store, rather than from some warehouse, such as a remote suburban area.

In short, the front warehouse is arranged at a place close to the user, for example, the relevant product purchased by the user may come from a front warehouse arranged in a nearby community, so that the user can be distributed to the home in a short time after ordering.

Illustratively, as shown in fig. 1, the fresh electric commerce platform adopts a two-stage distributed storage system of a city sorting center (a city total bin) + a front bin. For example, a plurality of city sorting centers are established, and a plurality of front-end bins are established in a plurality of communities or business circles according to factors such as order density, so that the product quality and the distribution speed are guaranteed.

User preferences: i.e. the degree of preference of the product by the user.

As an example, taking the fresh electric commerce platform as an example, the user preference refers to the user's preference degree for various fresh products.

User portrait: is a visual representation of data associated with a user, and in short, a user representation is a tagmentation of user information. In one possible implementation, the user representation may be a biased or statistically significant result of mining historical data for multiple dimensions. Namely, the user portrait can be constructed by mining historical data; the created portrait may be updated according to new data, which is not specifically limited in the embodiment of the present application.

Illustratively, a user representation includes, but is not limited to, user base data and user behavior data. The basic data of the user includes, but is not limited to, age, gender, occupation, constellation, blood type, graduation institution, area, income condition, and the like; the user behavior data includes, but is not limited to, click behavior, browsing behavior, purchasing behavior, and the like, and the embodiment of the present application does not limit the specific content included in the user basic data and the user behavior data.

Fig. 2 is a schematic diagram of an implementation environment related to an information pushing method provided in an embodiment of the present application.

Referring to fig. 2, the implementation environment includes: a user terminal 201 and a server 202. For example, the server 202 may be a fresh e-commerce platform, which is not particularly limited in this embodiment of the present application.

In one possible implementation, the types of the user terminal 201 include, but are not limited to: mobile terminals and fixed terminals. As an example, mobile terminals include, but are not limited to: smart phones, tablet computers, notebook computers, electronic readers, MP3 players (Moving Picture Experts Group Audio Layer III, Moving Picture Experts compress standard Audio Layer 3), MP4 players (Moving Picture Experts Group Audio Layer IV, Moving Picture Experts compress standard Audio Layer 4), etc.; the fixed terminal includes, but is not limited to, a desktop computer, which is not particularly limited in the embodiments of the present application.

Exemplarily, fig. 2 is only illustrated by taking the user terminal 201 as a smart phone. And the server 202 is used to provide background services for the user terminal 201. The server 102 may be an independent server, or may be a server cluster composed of a plurality of servers, which is not limited in this embodiment of the present application. In addition, an adaptive application is usually installed on the user terminal 202, so that the user can browse the products and place orders through the adaptive application conveniently.

Based on the above implementation environment, a user can place an order through the user terminal 201, and the related order is usually distributed to the micro-warehouse near the user by the server 202, for example, the related fresh product purchased by the user may come from a micro-warehouse arranged in the community near the user, so that the product purchased by the user in a short time after placing the order can be distributed to the home.

As is known, collaborative filtering recommendation usually only considers user history data (such as user history behavior), but it is difficult to make accurate recommendation if user related data (such as user interest information) is missing; the content-based recommendation method has a problem that the recommended content is single. In short, for a new user, effective recommendation cannot be performed due to the data sparsity problem, the cold start problem, the user-related data loss problem and the like in the related technology; for old users, it is difficult to find new points of interest for the users, and only products similar to the existing interests of the users can be recommended. Namely, the related art has the defect that product recommendation cannot be effectively performed, and therefore the pushed information is weak in pertinence, and the effect is poor.

In order to solve the above problems, an embodiment of the present application provides a method for pushing personalized information of a user preference product and a hot-sell product based on a micro-warehouse. The method uses the micro-bin as granularity and combines with the user portrait data, can well solve the problems of cold start and the like caused by less data of a new user, and cannot perform personalized recommendation to the new user; meanwhile, the problem that products are pushed to old users only singly is solved, the user opening rate and the user conversion rate of related application programs can be remarkably improved, and the effect is good.

The following embodiments are provided to describe the information pushing method in detail.

Fig. 3 is a flowchart of an information pushing method according to an embodiment of the present application. Referring to fig. 3, a method flow provided by the embodiment of the present application includes:

301. and determining a target preposed bin corresponding to the user, wherein the target preposed bin is the preposed bin of which the user has product transaction behaviors.

302. And obtaining M user preference products of the target front bin, wherein the M user preference products are determined according to the user preference products of each user under the target front bin, and M is a positive integer.

303. N kinds of hot selling products of the target preposed bin are obtained, the N kinds of hot selling products are determined according to the product transaction amount under the target preposed bin, and N is a positive integer.

304. Determining products to be pushed for the user according to the M types of user preference products and the N types of hot sales products, generating pushing information of the products to be pushed, and sending the generated pushing information to the user.

The method provided by the embodiment of the application comprises the steps of firstly determining a target preposed bin corresponding to a user, wherein the target preposed bin is a preposed bin of which the user has a product transaction behavior; then, obtaining M user preference products and N hot sale products of the target preposed bin, wherein the M user preference products are determined according to the user preference products of each user under the target preposed bin, the N hot sale products are determined according to the product transaction amount under the target preposed bin, and M and N are positive integers; and then, determining products to be pushed for the user according to the M types of user preference products and the N types of hot sales products, generating pushing information of the products to be pushed, and sending the generated pushing information to the user. Based on the above description, the information push is realized based on the user preference product and the hot sales product of the front bin associated with the user, and the method solves the problem that the product recommendation cannot be effectively carried out, so that the information push to the user is strong in pertinence and good in effect.

preprocessing the acquired user basic data and the user behavior data;

accumulating the quantity of the same products preferred by different users;

In one possible implementation, the method further includes:

acquiring a final preference score of each product in the target products;

acquiring preference weight values of the M types of user preference products;

acquiring hot selling weight values of the N hot selling products;

All the above optional technical solutions may be combined arbitrarily to form the optional embodiments of the present disclosure, and are not described herein again.

Fig. 4 is a flowchart of an information pushing method according to an embodiment of the present application. Exemplarily, taking an execution subject of the method as a user terminal and a fresh electric business platform as an example, referring to fig. 4, a method flow provided by the embodiment of the present application includes:

401. and the fresh electricity commercial platform determines a target user set to be subjected to information push.

The information push method is used for screening target crowds from massive push crowds, completing user filtering, and executing the information push method provided by the embodiment of the application aiming at the screened target crowds. The screened target group is also referred to as a target user set herein.

In a possible implementation manner, the embodiment of the application can filter the target user set through the user portrait data. Illustratively, the target user set may include a first class of users and a second class of users, that is, the first class of users and the second class of users constitute the target user set to be subjected to information push.

In the embodiment of the application, the first class of users refers to new users, and the second class of users refers to old users meeting specified conditions. Illustratively, the prescribed condition may be an old user who is not interested in historical pushes. In short, the fresh e-commerce platform can screen out old users carrying tags of 'new users' and tags of 'uninteresting to historical push' from a mass push crowd through user portrait data.

As an example, an old user who is not interested in the history push may be determined by the user not performing any operation on the history push, for example, the user does not perform an operation behavior such as clicking, purchasing or browsing, which is not specifically limited in this embodiment of the application.

Illustratively, new and old users may define the partition by the following dimensions. For a point in time, as user behavior occurs, the state of a new/old user of a user changes. For example, the user who generates the first order placing behavior is a new user, and the user who generates the second and later order placing behaviors is an old user; for another example, the user who opens the relevant application for the first time is a new user, and the user who opens the new relevant application for the second time or later is an old user. For a period of time, as the observation time window changes, the state of a new/old user of one user changes. For example, the user who appears for the first time during the observation period and generates the order placing behavior is the new user, otherwise, the user is the old user.

It should be noted that, for each user in the target user set, the information push can be performed by performing the following steps 402 to 405.

402. For each user in the target user set, the fresh electric commercial platform determines a target front-end bin corresponding to the user.

In the embodiment of the application, the micro-bin information corresponding to the user can be acquired through identification Information (ID) of the user. Illustratively, the user ID may be a user name, a phone number, etc. that the user fills in when registering on the associated application. And the micro-bin information may be ID information of the micro-bin.

In the embodiment of the present application, the target front bin refers to a micro bin that the user has made an order, that is, the target front bin is a front bin where the user has a product transaction behavior. As one example, a target pre-bin may refer to a pre-bin that the user has made the most order over a past period of time, such as a target pre-bin may be a pre-bin that the user has made the most order within 90 days of history.

Exemplary, the determination of the target leading bin includes, but is not limited to: the stored log is analyzed and processed to obtain an order record of the user, and then the target front-end bin corresponding to the user is determined according to the order record of the user.

403. And the fresh E-commerce platform acquires the M user preference products of the target front bin.

In the embodiment of the application, the user preference product of the front bin is determined according to the user preference product of each user under the front bin, namely, the user preference product of each user which is singled under the front bin. And M is a positive integer, namely obtaining a user preference product TOPM of the target front bin. Illustratively, the value of M may be 10, i.e. the user preference product TOP10 for acquiring the target front bin.

It should be noted that, in order to generate M types of user-preferred products of the target front-end bin, the user-preferred products of each user are generated first, where fig. 5 shows a process flow of determining the user-preferred product of any one user; and then, as shown in fig. 6 or fig. 7, through the association between the user and the micro-bin, the aggregation sorting finally generates M user preference products for the front bin.

In detail, for any user under any micro-bin, the process flow shown in fig. 5 includes, but is not limited to, the following steps:

4031. and acquiring the user basic data and the user behavior data of the user.

As previously described, user base data and user behavior data are both included in the user representation data. The basic data of the user includes, but is not limited to, age, gender, occupation, constellation, blood type, graduation institution, area, income condition, and the like; user behavior data includes, but is not limited to, click behavior, browse behavior, purchase behavior, and the like.

4032. And preprocessing the acquired user basic data and user behavior data.

In one possible implementation, the pre-processing includes, but is not limited to, data cleansing and data transformation.

The data cleaning is a procedure for finding and correcting recognizable errors in the data, and the step of cleaning is performed by selecting a proper method aiming at obvious error values, missing values, abnormal values and suspicious data found in the data examination process, so that dirty data are changed into clean data, and subsequent statistical analysis is facilitated to obtain a reliable conclusion. Of course, the data cleaning may further include deleting the duplicate records, checking data consistency, and the like, which is not specifically limited in this embodiment of the application. And the data conversion is mainly used for carrying out normalization operation on the data to enable the data to become a data source meeting the data mining requirement.

4033. And performing feature extraction on the preprocessed user data, inputting the generated feature data into a user preference model, and acquiring an output result of the user preference model, wherein the output result gives a preference weight value reflecting the user to a corresponding product type.

In the embodiment of the application, the preference model of the user gives the preference weight value of the user to the corresponding product category. The preference weight value reflects the preference degree of the user to the corresponding product type, and the greater the preference weight value is, the greater the preference degree of the user to the corresponding product type is indicated. For example, a value of the preference weight value may be 1 to 10 points, which is not specifically limited in this embodiment of the application.

It should be noted that, taking the target pre-bin as an example, for each user under the target pre-bin, the user preferred product of the corresponding user can be determined through the above steps 4031 to 4033.

In one possible implementation, the user preference model may be a linear model. For example, an initial linear model between the user representation data and the user preferred product is constructed; training an initial linear model based on training sample data, namely optimizing linear parameters; and adopting the optimized linear parameters as the linear parameters of the final linear model. The user portrait data includes, but is not limited to, user basic number and user behavior data.

In another possible implementation, the user preference model may also be a non-linear deep learning model. For this case, one possible training process for the user preference model is as follows: acquiring training sample data, wherein the training sample data comprises user portrait data of a plurality of users collected in advance and actual preference product types marked to the users; inputting user portrait data in training sample data into a deep learning model; determining whether the predicted preferred product type output by the deep learning model is matched with the marked preferred product type or not based on the target loss function; and if the predicted preferred product type is not matched with the actual preferred product type, iteratively updating the network parameters of the deep learning model repeatedly and circularly until the model converges to obtain the user preference model. Wherein, the target loss function may be a cross entropy loss function; the condition of whether the matching is performed or not may be that the error is within a preset range, and the model convergence condition may be that the prediction accuracy of the model reaches more than 95%, which is not specifically limited in the embodiment of the present application.

In detail, the process flow shown in fig. 6 includes, but is not limited to, the following steps:

4034. and determining each user with product transaction behavior in the target preposed bin according to the identification information of the target preposed bin.

For this step, the individual users that have been singled under the target micro-bin are matched according to the ID information of the target pre-bin.

4035. And according to the dimension of the front bin, aggregating the user preference products of each user under the target front bin to obtain the user preference products matched with the target front bin.

It should be noted that the user preference product of each user may be one or more, and this is not specifically limited in this embodiment of the application.

For this step, the user preference products of the user are aggregated at the granularity of micro-bins. Illustratively, assuming 100 users under the target front bin, the user-preferred products of the 100 users that were singled under the target micro-bin are aggregated.

4036. Accumulating the quantity of the same products preferred by different users; and sorting the user preference products matched with the target front bin according to the sequence of the accumulated product quantity from large to small.

The types of products favored by users under the target micro-warehouse may be the same or different, illustratively, the embodiment of the application accumulates the number of times of statistics on similar products favored by different users to obtain the number of products, namely, when the same products favored by different users occur, the corresponding number of products is accumulated; and then sorting the user preference products matched with the target front bin according to the sequence of the obtained product quantity from large to small.

As an example, taking the number of users in a certain micro bin as 3 as an example, assuming that the product preferred by the user a is an apple, the product preferred by the user B is an apple, and the product preferred by the user C is a pear, the number of products corresponding to the apple is 2, and the number of products corresponding to the pear is 1, so that the user prefers that the apples are arranged in front of the pears.

4037. And taking the user preference products ranked at the top M as the M user preference products of the target front bin.

In addition, in addition to the determination of M user preferred products by the way of accumulating the number of products as shown in fig. 6, the M user preferred products may also be determined by the way of accumulating the preference weight values as shown in fig. 7, and the processing flow includes, but is not limited to, the following steps:

4038. and determining each user with product transaction behavior in the target preposed bin according to the identification information of the target preposed bin.

4039. And according to the dimension of the front bin, aggregating the user preference products of each user under the target front bin to obtain the user preference products matched with the target front bin.

4040. Accumulating the same products preferred by different users according to the preference weight values; and sequencing the user preference products matched with the target front bin according to the sequence of the accumulated preference weight values from large to small.

The types of products favored by users under the target micro-warehouse may be the same or different, and exemplarily, the preference weight values of similar products favored by different users are accumulated to obtain the accumulated preference weight values; and then sorting the user preference products matched with the target front bin according to the sequence of the accumulated preference weight values from large to small.

4041. And taking the user preference products ranked at the top M as the M user preference products of the target front bin.

In another possible implementation manner, after the M user preference products are obtained, new preference weight values are given to the M user preference products according to the above ordering. The value of the new preference weight value may be 1 to 10 points, which is not specifically limited in the embodiment of the present application. Generally, the more top ranked user preference products have higher preference weights.

The process of determining the micro-bin preference shown in step 403 above is described below by way of an example.

For each user, the user preference product may be one or more and have a preference weight value. Wherein, table 1 shows the products preferred by different users and the corresponding preference weight values (only one user is used for illustration). Illustratively, the preference weight value in table 1 is 1 to 10 points.

TABLE 1

User ID	User preference product	Preference weight value
			1	Apple and pear	Apple: 10, pear: 9
2	Milk
			3	Pear (pear)
4	Banana
			5	Apple (Malus pumila)
6	Apple (Malus pumila)
			7	Milk
8	Apple (Malus pumila)
			9	Milk
10	Pear (pear)

Further, table 2 gives the correspondence between the user and the corresponding target pre-bin. Illustratively, the front bins given in table 2 are the front bins that were most singled by the user within 90 days of the history.

TABLE 2

Further, as shown in table 3, the user preferred products under the front bin W1 shown in table 2 are sorted in order of increasing accumulated product quantity (and also in order of decreasing accumulated preference weight value).

TABLE 3

Front bin ID	User preference product	Cumulative product quantity	Sorting
				W1	Apple (Malus pumila)	4	1
W1	Milk	3	2
				W1	Pear (pear)	2	3
W1	Banana	1	4
				…	…	…	…

Then, the preference weight values are assigned again to the M user preference products intercepted from table 3 in the above-mentioned order, see table 4 below.

TABLE 4

Front bin ID	User preference product	Reassigned preference weight values
			W1	Apple (Malus pumila)	10
W1	Milk	9
			W1	Pear (pear)	8
W1	Banana	7
			…	…	…

404. The fresh electric commerce platform obtains N hot-selling products of the target front-end bin.

In the embodiment of the application, the hot selling products are also called hot selling products and are determined according to the product transaction amount under the front bin, namely, the sales amount of each product sold by the front bin is determined, and the greater the sales amount of a certain product is, the hotter the product is indicated.

And N is a positive integer, namely obtaining a TOPN of the hot-pin product of the target front bin. The value of N may be the same as or different from the value of M, which is not specifically limited in this embodiment of the present application. Illustratively, the value of N may be 10, i.e., the hot-pin product TOP10 of the target leading bin is obtained.

It should be noted that there is a certain difference between the hot-selling products and the user-preferred products mentioned in the embodiments of the present application. Generally, the user preference product and the hot-sell product will mostly coincide, but in some special cases, there will be a difference between the two. For example, some hot-selling products may belong to the preferences of a small number of users, but the purchase amount of the part of users is large, so that the user preference product and the hot-selling product are considered together in the embodiment of the present application.

In a possible implementation manner, when acquiring the N types of hot-selling products of the target front-end bin, after analyzing the stored log, acquiring the ordering record of the user in the target front-end bin, and then counting the sales volume of each product type through the ordering record in the target front-end bin, thereby taking the products with the sales volume sorted in the top N positions as the N types of hot-selling products of the target front-end bin.

For example, after acquiring the N kinds of hot-sell products, the embodiments of the present application may further assign hot-sell weight values to the N kinds of hot-sell products according to the above ordering. The value of the hot-sell weight value may be 1 to 10 points, which is not specifically limited in the embodiment of the present application. Generally, the higher the ranking, the higher the weight of the hot-pin product.

405. The fresh E-commerce platform determines products to be pushed for the user according to the M types of user preference products and the N types of hot sales products, generates pushing information of the products to be pushed and sends the generated pushing information to the user.

For the new user and the old user, as shown in fig. 8, the embodiment of the present application differs in determining the product to be pushed, specifically referring to steps 4051 and 4052 described below.

4051. When the user is a new user, determining products to be pushed for the user according to the M user preference products and the N hot sales products, including but not limited to:

acquiring the final favorite score of each of the M user preference products and the N hot sales products; according to the obtained final preference score, sequencing M user preference products and N hot sales products in a descending order; and taking the products ranked at the top X as the products to be pushed for the user, wherein X is a positive integer.

For example, taking the values of M and N as 10 and the value of X as 3 as an example, the 20 products are sorted in the order of the final preference scores from large to small, and the product sorted in the top 3 bits is taken as the product to be pushed for the user.

4052. When the user is an old user who is not interested in historical pushing, products to be pushed for the user are determined according to the M types of user preference products and the N types of hot sales products, and the method comprises the following steps:

step a, determining a first class of products and a second class of products corresponding to a user, wherein the first class of products are historical push products, and the second class of products are products which are contrary to the user preference of the user.

In the embodiment of the application, the first type of products refers to historical push products which are already pushed by old users, and the second type of products refers to products which are disliked by the users. Illustratively, the product that the user dislikes may be obtained by analyzing the user portrait data or by analyzing a stored log, which is not particularly limited in the embodiment of the present application.

And b, filtering the first type of products and the second type of products in the M types of user preference products and the N types of hot sales products to obtain target products.

The step is to filter the product. Wherein the target product is M user preferred products + N hot-sold products-historical pushed products-products that the user dislikes.

Step c, obtaining the final favorite score of each product in the target products; and sequencing the target products according to the obtained final preference score in a descending order, and taking the products sequenced at the front X positions as products to be pushed for the user.

In the embodiment of the present application, the obtaining of the final preference score includes, but is not limited to, the following ways: acquiring preference weight values of M types of user preference products and hot sales weight values of N types of hot sales products; acquiring a first coefficient of a user preference product and a second coefficient of a hot-selling product; and determining a final preference score based on the obtained preference weight value and the obtained hot sales weight value and the obtained first coefficient and second coefficient.

The above implementation is expressed by mathematical formula, i.e. y-q 1 w1+ q2 w 2; where q1 denotes a first coefficient, w1 denotes a preference weight value (newly assigned preference weight value), w2 denotes a hotbox weight value, q2 denotes a second coefficient, and y denotes a final taste score. For example, the values of the first coefficient and the second coefficient may be respectively 0.5, which is not specifically limited in this embodiment of the application.

As an example, if the M user preference products and the N hot-selling products each include an apple, and the preference weight value corresponding to the apple is 10, and the hot-selling weight value is 8, the final preference score y corresponding to the apple is 10 × 0.5+8 × 0.5 — 9. The final preference score may be rounded up or naturally rounded up of the calculation result, which is not specifically limited in this embodiment of the application.

406. And the user terminal displays a push message sent by the fresh E-commerce platform.

In the embodiment of the application, the user terminal can display the push message sent by the fresh electronic commerce platform through the fresh application program. For example, the push message may be displayed by the fresh class application program when the user opens the fresh class application program, or the push message may be partially displayed by the system message notification bar when the fresh class application program is in the background running state. Illustratively, the push message includes, but is not limited to, a teletext message, which is also not specifically limited in the embodiments of the present application.

The method provided by the embodiment of the application has at least the following beneficial effects:

the information pushing method and the device have the advantages that the information pushing is achieved based on the user preference products and the hot sales products of the front bin associated with the user, the problem that the product recommendation cannot be effectively carried out is solved, the pertinence of the information pushed to the user is strong, and the effect is good. In another expression mode, the embodiment of the application provides a method for pushing personalized information of user preference products and hot-sell products based on micro-bins. The method uses the micro-bin as granularity and combines with the user portrait data, can well solve the problems of cold start and the like caused by less data of a new user, and cannot perform personalized recommendation to the new user; meanwhile, the problem that products are pushed to old users and lack of diversity and singleness is solved, the user opening rate and the user conversion rate of the related application programs can be remarkably improved, and the effect is good.

Fig. 9 is a schematic structural diagram of an information pushing apparatus according to an embodiment of the present application. Referring to fig. 9, the apparatus includes:

a first determining module 901, configured to determine a target pre-bin corresponding to a user, where the target pre-bin is a pre-bin where a product transaction behavior exists for the user;

a first obtaining module 902, configured to obtain M user preference products of the target pre-bin, where the M user preference products are determined according to user preference products of users under the target pre-bin, and M is a positive integer;

a second obtaining module 903, configured to obtain N hot-sell products of the target pre-bin, where the N hot-sell products are determined according to a product transaction amount under the target pre-bin, and N is a positive integer;

a second determining module 904 configured to determine products to be pushed for the user according to the M user preference products and the N hot-sell products;

a pushing module 905 configured to generate pushing information of the product to be pushed, and send the pushing information to the user.

According to the device provided by the embodiment of the application, for each user in a target user set, a target front-end bin corresponding to the user is determined at first, wherein the target front-end bin is a front-end bin in which product transaction behaviors exist in the user; then, obtaining M user preference products and N hot sale products of the target preposed bin, wherein the M user preference products are determined according to the user preference products of each user under the target preposed bin, the N hot sale products are determined according to the product transaction amount under the target preposed bin, and M and N are positive integers; and then, determining products to be pushed for the user according to the M types of user preference products and the N types of hot sales products, generating pushing information of the products to be pushed, and sending the generated pushing information to the user. Based on the above description, the information push is realized based on the user preference product and the hot sales product of the front bin associated with the user, and the method solves the problem that the product recommendation cannot be effectively carried out, so that the information push to the user is strong in pertinence and good in effect.

In one possible implementation, the apparatus further includes:

a first determining module, further configured to determine a target pre-bin corresponding to each user in the set of target users;

It should be noted that: in the information pushing apparatus provided in the foregoing embodiment, only the division of the function modules is illustrated when information is pushed, and in practical applications, the function distribution may be completed by different function modules according to needs, that is, the internal structure of the apparatus is divided into different function modules to complete all or part of the functions described above. In addition, the information pushing apparatus and the information pushing method provided by the above embodiments belong to the same concept, and specific implementation processes thereof are detailed in the method embodiments and are not described herein again.

Fig. 10 is a schematic structural diagram of an information pushing apparatus according to an embodiment of the present application. Illustratively, the information pushing device may be embodied as a server. The server 1000 may generate a relatively large difference due to different configurations or performances, and may include one or more processors (CPUs) 1001 and one or more memories 1002, where the memory 1002 stores at least one instruction, and the at least one instruction is loaded and executed by the processors 1001 to implement the information pushing method provided by each method embodiment. Of course, the server may also have components such as a wired or wireless network interface, a keyboard, and an input/output interface, so as to perform input/output, and the server may also include other components for implementing the functions of the device, which are not described herein again.

In an exemplary embodiment, a computer-readable storage medium, such as a memory, including instructions executable by a processor in a terminal to perform the information pushing method in the above-described embodiments is also provided. For example, the computer readable storage medium may be a ROM, a Random Access Memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like.

It will be understood by those skilled in the art that all or part of the steps for implementing the above embodiments may be implemented by hardware, or may be implemented by a program instructing relevant hardware, where the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.

The above description is only exemplary of the present application and should not be taken as limiting, as any modification, equivalent replacement, or improvement made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. An information pushing method, characterized in that the method comprises:

2. The method of claim 1, wherein said obtaining M user-preferred products for the target pre-bin comprises:

3. The method of claim 2, wherein the obtaining user-preferred products for respective users under the target pre-bin comprises:

preprocessing the acquired user basic data and the user behavior data;

4. The method of claim 1, further comprising:

determining a first class of users and a second class of users according to user portrait data, wherein the first class of users and the second class of users form a target user set to be subjected to information pushing;

5. The method of claim 4, wherein when the user is the first class of user, the determining the products to be pushed for the user according to the M user preferred products and the N hot-sell products comprises:

6. The method of claim 4, wherein when the user is the second type of user, the determining the products to be pushed for the user according to the M user preferred products and the N hot-sell products comprises:

acquiring a final preference score of each product in the target products;

7. The method of claim 5 or 6, wherein the obtaining of the final preference score comprises:

acquiring preference weight values of the M types of user preference products;

acquiring hot selling weight values of the N hot selling products;

8. An information pushing apparatus, characterized in that the apparatus comprises:

9. A computer-readable storage medium, wherein at least one instruction is stored in the storage medium, and is loaded and executed by a processor to implement the information pushing method according to any one of claims 1 to 7.

10. An information pushing apparatus, characterized in that the apparatus comprises a processor and a memory, wherein the memory stores at least one instruction, and the at least one instruction is loaded and executed by the processor to implement the information pushing method according to any one of claims 1 to 7.