WO2018014764A1

WO2018014764A1 - Method and device for selecting commodity object, and determining model and usage popularity

Info

Publication number: WO2018014764A1
Application number: PCT/CN2017/092588
Authority: WO
Inventors: 叶舟; 王瑜; 陈凡; 杨洋; 董昭萍; 钱倩; 王吉能
Original assignee: 阿里巴巴集团控股有限公司; 叶舟; 王瑜; 陈凡; 杨洋; 董昭萍; 钱倩; 王吉能
Priority date: 2016-07-19
Filing date: 2017-07-12
Publication date: 2018-01-25
Also published as: TW201816685A; CN107632989A; CN107632989B

Abstract

Provided are a method and device for selecting a commodity object, and determining a model and usage popularity. The invention can automatically select, from a large number of objects, and on the basis of a keyword associated with a commodity object and inputted by a user (10) and a set object identification model, at least one commodity object as an object meeting a requirement of the user (10), thereby significantly improving efficiency of selecting commodity objects, reducing costs incurred in performing stocktaking by personnel, and improving operation efficiency.

Description

Method and device for selecting commodity object, model determination and using heat determination

Technical field

The present application relates to the field of Internet technologies, and in particular, to a commodity object selection, model determination, and usage heat determination method and apparatus.

Background technique

In order to improve the transaction performance of commodity objects in the e-commerce system, the e-commerce system often establishes various new channels to increase the exposure of the product objects, for example, establishing various special spike activity theme channels, or the main product object tonality. Theme channels and more.

These channels will encounter the problem of inventory when they are just established, that is, how to delineate the product objects suitable for the channel, in order to maximize the channel turnover. Specifically, at present, in order to solve the above problem, the corresponding product object is often selected manually for each new channel, that is, the operator subjectively selects the product object that meets the required theme of each new channel according to the human experience.

However, since it takes a lot of time to manually select a product object, the selection efficiency of the product object is very low.

Summary of the invention

The embodiment of the present application provides a method and device for selecting a product object, determining a model, and determining a usage heat, so as to solve the problem of inefficiency existing in the existing product object selection mode.

In one aspect, the embodiment of the present application provides a method for selecting a commodity object, including:

Receiving a commodity object keyword sent by the user terminal;

Obtaining a primary selection object that matches the subject object of the commodity object;

Determining the object usage heat of each primary object in the second time period based on the set object recognition model and the use feature data of each of the preliminary objects in the first time period; wherein the object recognition model is used for training Characterizing the use of the product object in the first time period, and the product object The object in the second time period uses an identification model of the relationship between the heats; the first time period is a previous specified time period of the second time period;

And selecting at least one commodity object from the primary selection objects as the commodity object matching the commodity object keyword in the second time period according to the object usage heat of each of the primary selection objects in the second time period.

Optionally, selecting at least one commodity object from the primary selection object as the commodity object in the second time period that matches the commodity object keyword includes:

Selecting at least one object from the primary selection object uses a commodity object having a heat not lower than a set heat threshold as a commodity object in the second time period that matches the commodity object keyword.

Optionally, before determining, according to the set object identification model and the use feature data of each of the preliminary objects in the first time period, before the object usage heat of each of the preliminary objects in the second time period, the method further includes:

Receiving object recognition model training sample data sent by the user terminal, wherein the object recognition model training sample data includes basic feature data of each object recognition model sample object, and the basic feature data of each object recognition model sample object includes Using the feature data of the object recognition model sample object in each third time period, and the object use heat of the object recognition model sample object in each corresponding fourth time period; the third time period is the corresponding fourth The previous specified time period of the time period;

Determining, according to the basic feature data of each object recognition model sample object, a pre-established relationship between the use feature data for predicting the product object in the first time period and the object use heat of the product object in the second time period The initial object recognition model is trained to obtain the object recognition model.

Optionally, before the at least one product object is selected from the primary selection object as the commodity object in the second time period that matches the product object keyword, the method further includes:

Determining each primary selection object according to the set category identification model and the category corresponding to each primary selection object in the category usage heat of one or more historical synchronization time periods corresponding to the second time period The category uses heat in the category of the second time period; wherein the category recognition model is trained Between the use of the category of the commodity object category in the second time period, and the heat usage of the category of the commodity object category in the one or more historical time periods corresponding to the second time period Identification model of the relationship;

For each primary selection object, the hotspot of the second time period is used to update the object usage heat in the second time period according to the category corresponding to the primary selection object.

Optionally, determining, according to the set category identification model, and the category corresponding to each primary selection object, the heat usage of the category of one or more historical synchronization time periods corresponding to the second time period Before the category corresponding to the primary selection object uses the heat in the category of the second time period, the method further includes:

Receiving the category identification model training sample data sent by the user terminal, wherein the category identification model training sample data includes basic feature data of various target recognition model sample objects, and each category identifies the model sample object basis The feature data includes the category usage heat of the category corresponding to the category identification model sample object in each fifth time period, and the category corresponding to the category identification model sample object corresponds to each fifth time period. The use of heat in one or more historical time periods;

According to the basic feature data of the sample objects of the various types of mesh recognition models, the pre-established use of the category for predicting the commodity object category in the second time period is used, and the commodity object category corresponds to the second time period. The category of one or more historical contemporaneous time periods is trained using an initial category recognition model of the relationship between the heats to obtain the category identification model.

The historical time period of each time period refers to a historical time period that is on the same calendar day or lunar day as the time period and corresponds to the time period.

Optionally, for any primary selection object, the object in the second time period is used in the category according to the category corresponding to the primary selection object, and the object of the primary selection object is in the second time period. Before using the heat to update, the method further includes:

If it is determined that the second time period is a specific time period, if it is determined that the category corresponding to the primary selection object is a category that matches a specific time period corresponding to the second time period, according to the set coefficient And increasing the category usage heat of the category corresponding to the primary selection object in the second time period.

Optionally, at least one commodity object is selected from the preliminary objects as the second time period, Before the commodity object matching the commodity object keyword, the method further includes:

Determining the similarity between the primary selected objects according to the similarity between the titles corresponding to the primary objects;

For each set of objects consisting of at least one primary selected object whose degree of similarity is not lower than the first similarity threshold, one object in the set of objects is retained, and other objects are deleted.

Optionally, before acquiring the primary selected object that matches the commodity object keyword, the method further includes:

Determining, according to the set sample corpus, at least one sample word having a similarity between the product object subject and the second similarity threshold based on the set sample corpus;

The determined sample words are used as the final desired commodity object keywords.

The object recognition model is a regression model; the category recognition model is a linear model.

On the other hand, the embodiment of the present application provides another method for selecting a commodity object, including:

Receiving a product object keyword input by the user, and sending the product object keyword to the server;

Receiving the product object information returned by the server according to the product object keyword, and using the product object corresponding to the product object information as the product object matching the product object keyword in the second time period;

The product object information is used by the server according to the set object recognition model and the primary selection objects that match the product object topic in the first time period, and the product data from the product. Corresponding information of the commodity object selected in the primary object matched by the object keyword; the first time period is a previous specified time period of the second time period;

The set object recognition model is used to characterize the relationship between the use feature data of the product object in the first time period and the object use heat of the product object in the second time period.

Optionally, before receiving the product object keyword input by the user, the method further includes:

Receiving object recognition model training sample data input by the user, wherein the object recognition model training sample data includes basic feature data of each object recognition model sample object, and each object recognition The base feature data of the different model sample object includes the use feature data of the object recognition model sample object in each third time period, and the object use heat of the object recognition model sample object in each corresponding fourth time period; The third time period is a previous specified time period of the corresponding fourth time period;

Transmitting the object recognition model training sample data to a server, wherein the server identifies the basic feature data of each object recognition model sample object included in the sample data according to the object recognition model, and the pre-established product object for predicting The initial object recognition model using the relationship between the feature data and the object use heat in the second time period in the first time period is trained to obtain the set object recognition model.

In another aspect, the embodiment of the present application provides a method for determining a model, including:

And according to the basic feature data of each object recognition model sample object included in the object recognition model training sample data, the pre-established use feature data for predicting the product object in the first time period, and the commodity object in the second time The object of the segment is trained using the initial object recognition model of the association relationship between the heats, and is obtained between the use feature data for characterizing the commodity object in the first time period and the object use heat of the commodity object in the second time period. An object recognition model of the association relationship; the first time period is a previous specified time period of the second time period.

The usage characteristic data includes at least one or more of the number of browsing times, the number of collections, the number of purchases, the number of transactions, the number of comments, and the number of searches; the usage heat of the object includes at least volume, turnover, And any one or more of the transaction conversion rates.

In another aspect, the embodiment of the present application provides another method for determining a model, including:

Receiving, by the user terminal, a category identification model training sample data, wherein the category identification module The type training sample data includes basic feature data of various target recognition model sample objects, and the basic feature data of each category recognition model sample object includes the category corresponding to the category identification model sample object at each first time The category of the segment uses the heat, and the category corresponding to the category identification model sample object uses the heat in the category of one or more historical time periods corresponding to the respective first time periods;

Generating the basic feature data of the sample objects of the various types of mesh recognition models included in the sample data according to the category identification model, and using the pre-established category for predicting the use of the commodity object category in the second time period, and the commodity object The category is trained in an initial category recognition model of the association relationship between the categories of heat usage using one or more historical time periods corresponding to the second time period, and is obtained in the second category for characterizing the commodity object. The category of the time period uses a heat classification, a category recognition model that is related to the heat usage of the category of the commodity object category in one or more historical time periods corresponding to the second time period.

The heat usage of the category includes at least one or more of a volume, a turnover, and a transaction conversion rate.

In another aspect, the embodiment of the present application further provides a method for determining a heat usage, including:

Obtaining usage characteristic data of each commodity object in a first time period;

Determining each item based on the relationship between the use feature data of the product object in the first time period and the object use heat of the product object in the second time period, and the use feature data of each product object in the first time period The object uses heat in the second time period;

The first time period is a previous specified time period of the second time period; and the association relationship is based on usage characteristic data of each sample object in each third time period and each sample object is in each corresponding fourth time The object of the segment is established using the heat; the third time period is the previous specified time period of the corresponding fourth time period.

In another aspect, the embodiment of the present application further provides a commodity object selection device, including:

a keyword receiving unit, configured to receive a commodity object keyword sent by the user terminal;

An object obtaining unit, configured to acquire a primary selection object that matches the keyword of the commodity object;

a heat determining unit for identifying a model based on the set object and each of the primary objects at the first time Using the feature data in the segment, determining the object usage heat of each primary object in the second time period; wherein the object recognition model is the training use characteristic data used to represent the commodity object in the first time period, and a recognition model of an association relationship between the objects of the commodity object in the second time period; the first time period is a previous specified time period of the second time period;

An object screening unit, configured to select at least one commodity object from the primary selection object as the second time period according to the object usage heat of each primary selection object in the second time period, and the subject matter of the commodity object Matching item object.

On the other hand, the embodiment of the present application further provides another commodity object selection device, including:

a keyword receiving unit, configured to receive a commodity object keyword input by the user;

a keyword sending unit, configured to send the commodity object keyword to a server;

An object information receiving unit, configured to receive commodity object information returned by the server according to the commodity object keyword;

An object determining unit, configured to determine a product object corresponding to the product object information as a product object that matches the product object keyword in a second time period;

In another aspect, the embodiment of the present application further provides a model determining apparatus, including:

a data receiving unit, configured to receive object recognition model training sample data sent by the user terminal, where the object recognition model training sample data includes basic feature data of each object recognition model sample object, and each object recognition model sample object The basic feature data includes the use feature data of the object recognition model sample object in each third time period, and the object use heat of the object recognition model sample object in each corresponding fourth time period; the third time The segment is the corresponding fourth time The previous specified time period of the segment;

a model training unit, configured to perform, according to the basic feature data of each object recognition model sample object included in the sample identification model training sample data, the pre-established use characteristic data for predicting the commodity object in the first time period, and The commodity object is trained in the initial object recognition model of the relationship between the objects using the heat in the second time period, and the use feature data for characterizing the commodity object in the first time period and the second time period of the commodity object are obtained. An object recognition model in which an object uses an association relationship between heats.

On the other hand, the embodiment of the present application further provides another model determining apparatus, including:

a data receiving unit, configured to receive the category identification model training sample data sent by the user terminal, wherein the category identification model training sample data includes basic feature data of various target recognition model sample objects, and each category Identifying the basic feature data of the model sample object includes the category usage heat of the category corresponding to the category identification model sample object in each first time period, and the category corresponding to the category identification model sample object in each category The usage of the category of one or more historical time periods corresponding to a period of time;

a model training unit, configured to: according to the category identification model, the basic feature data of the sample objects of the various types of mesh recognition models included in the sample data, and the pre-established categories for predicting the category of the commodity object in the second time period Training is performed using the initial category recognition model of the relationship between the heat and the category of the commodity object category in the one or more historical time periods corresponding to the second time period, and is used to characterize the commodity. The category of the object category in the second time period is the category of the association between the heat usage and the category usage heat of the commodity object category in one or more historical time periods corresponding to the second time period. Identify the model.

In a further aspect, the embodiment of the present application further provides a heat determining device, including:

a data obtaining unit, configured to acquire usage characteristic data of each commodity object in a first time period;

a heat determining unit, configured to use an association relationship between the use feature data of the product object in the first time period and the object use heat of the product object in the second time period, and use of each commodity object in the first time period Feature data, determining the object usage heat of each commodity object in the second time period;

The beneficial effects of the application are as follows:

The embodiment of the present application provides a method and device for selecting a product object, determining a model, and determining a usage heat, and automatically selecting at least one product object from a mass object based on a product object keyword input by the user and a set object recognition model. As the ultimate object to meet the needs of users, the efficiency of selecting goods objects is greatly improved, thereby reducing the cost of manual goods and improving operational efficiency.

DRAWINGS

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following drawings will be briefly described in the description of the embodiments. It is obvious that the drawings in the following description are only some embodiments of the present application, Those skilled in the art can also obtain other drawings based on these drawings without paying any creative work.

FIG. 1 is a schematic diagram of a possible application scenario of a method for selecting a commodity object according to Embodiment 1 of the present application;

FIG. 2 is a schematic flowchart diagram of a method for selecting a commodity object in the first embodiment of the present application;

FIG. 3 is a schematic flowchart diagram of a method for determining a model in the first embodiment of the present application;

FIG. 4 is a schematic flowchart diagram of another method for determining a model in the first embodiment of the present application;

FIG. 5 is a schematic flowchart diagram of a method for determining a heat usage according to Embodiment 1 of the present application;

FIG. 6 is a schematic diagram showing a possible structure of a product object selection device in Embodiment 2 of the present application;

FIG. 7 is a schematic diagram showing a possible structure of another commodity object selection device in the second embodiment of the present application;

FIG. 8 is a schematic structural diagram of a model determining apparatus in Embodiment 2 of the present application;

FIG. 9 is a schematic diagram showing a possible structure of another model determining apparatus in Embodiment 2 of the present application;

FIG. 10 is a schematic diagram showing a possible structure of a heat determining device according to Embodiment 2 of the present application.

detailed description

The present invention will be further described in detail with reference to the accompanying drawings, in which FIG. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments of the present application without departing from the inventive scope are the scope of the present application.

Embodiment 1:

In order to solve the problem of the inefficiency of the existing product object selection method, the first embodiment of the present application provides a method for selecting a commodity object, as shown in FIG. 1 , which is a possible method for selecting the commodity object. A schematic diagram of an application scenario, which may include, for example, a user terminal 11 and a server 12, where:

The user terminal 11 can receive the product object keyword input by the user 10 and send the product object topic word to the server 12; the server 12 can obtain the product object subject word according to the product object keyword sent by the user terminal 11 Matching the primary selection object, and based on the set object recognition model and the usage characteristic data of each primary selection object in the first time period, determining the object usage heat of each primary selection object in the second time period, and according to each primary selection The object uses the heat in the second time period, and selects at least one product object from the primary selection object as the commodity object in the second time period that matches the product object keyword; the user terminal 11 can receive The product object information of the at least one product object returned by the server 12 after selecting at least one product object, and the product object information Corresponding commodity object as a commodity object in the second time period that matches the commodity object keyword; wherein the set object recognition model can be used to characterize the use of the commodity object in the first time period The relationship between the feature data and the object usage heat of the commodity object in the second time period; the first time period is a previous specified time period of the second time period.

The user terminal 11 and the server 12 can perform a communication connection through a communication network, and the network can be a local area network, a wide area network, or the like. The user terminal 11 may be a terminal device such as a mobile phone, a tablet computer, a notebook computer, a personal computer, or even a client installed in the terminal device; the server 12 may be any server device capable of supporting processing operations such as screening of commodity objects. .

In other words, in the embodiment of the present application, at least one product object can be automatically selected from the mass object as the final object satisfying the user's demand based on the product object keyword input by the user and the set object recognition model, thereby The efficiency of selecting commodity objects is greatly improved, thereby reducing the cost of manual inventory and improving operational efficiency.

The product object selection method in the first embodiment of the present application will be exemplarily described with reference to FIG. 2 in conjunction with the application scenario shown in FIG. It should be noted that the above application scenarios are only for the purpose of facilitating understanding of the spirit and principle of the present application, and the embodiments of the present application are not limited in this respect. Rather, embodiments of the present application can be applied to any scenario that is applicable.

Specifically, as shown in FIG. 2, it is a possible flow diagram of the method for selecting a product object in the first embodiment of the present application. The method for selecting a product object may include the following steps:

Step 201: The user terminal receives the object recognition model training sample data input by the user, and sends the object recognition model training sample data to the server.

The object recognition model training sample data includes basic feature data of each object recognition model sample object, and the basic feature data of each object recognition model sample object includes the object recognition model sample object in each third time period. The use feature data within, and the object recognition model sample object uses heat in each corresponding fourth time period; the third time period is a previous specified time period of the corresponding fourth time period.

Optionally, the usage feature data of the commodity object such as each object recognition model sample object may include at least Any one or more of the number of times of browsing, the number of times of collection, the number of purchases (adding to the shopping cart), the number of transactions, the number of comments, and the number of searches; the object usage heat of each object object such as the object recognition model sample object may include at least a deal Any one or more of quantity, turnover, and transaction conversion rate. In addition, it is to be noted that the use feature data of the product object such as each object recognition model sample object can be acquired from the product object operation information of each e-commerce website, and the object usage heat of each product object such as the object recognition model sample object can be based on each The feature data of the product object such as the object recognition model sample object is calculated and will not be described here.

In addition, it should be noted that the third and fourth time periods may generally be historical time segments, that is, the basic feature data of each object recognition model sample object may generally be corresponding historical data; and, third, fourth The time period, and the time period of the specified time period can be flexibly set according to the actual situation, such as 1 day, 1 week, 1 month, etc. (usually a minimum of one day). In addition, the lengths of the third and fourth time periods may be the same or different, for example, the fourth time period may be one day, and the third time period corresponding to the fourth time period (ie, the previous designation of the fourth time period) The time period may be one month or one year, etc., or the fourth time period may be one month, and the third time period corresponding to the fourth time period may be one day, etc.; The three time periods may be the previous specified time period adjacent to the fourth time period, and may or may not be adjacent.

Step 202: The server receives the object recognition model training sample data sent by the user terminal, and according to the object recognition model, the basic feature data of each object recognition model sample object in the sample data is trained, and the pre-established initial object recognition model is trained to obtain the The required object recognition model.

The initial object recognition model is a recognition model for predicting an association relationship between the use feature data of the commodity object in the first time period and the object use heat of the commodity object in the second time period; the object recognition The model is a training model for characterizing the relationship between the use feature data of the commodity object in the first time period and the heat usage of the object object in the second time period, the first time period is The previous specified time period of the second time period.

In addition, similar to the foregoing descriptions regarding the third and fourth time periods, the sizes of the first and second time periods and the like may also be flexibly set according to actual conditions, and the lengths of the first and second time periods may be different. The same may be different (however, the size of the first time period may be the same as the third time period, and the size of the second time period may be the same as the fourth time period), which is not limited thereto.

Optionally, the object usage identification object model object and the object object use heat as the transaction volume, for example, the corresponding object recognition model can be obtained by the following steps:

A1: Establish an initial object recognition model related to volume.

Optionally, it is assumed that the third and fourth time periods can be set to 1 day; and it is assumed that y(t) can represent the volume of a certain commodity object on the date t, and the commodity object is represented by x1(t-1) The number of times of the date t-1, x2 (t-1) indicates the number of times the item is stored on the date t-1, and x3(t-1) indicates the number of times the item is purchased on the date t-1, etc. And, assuming that the initial object recognition model is a linear model; then the established initial object recognition model can be expressed as y(t)=a(1)*x1(t-1)+a(2)*x2(t-1) +a(3)*x3(t-1)+..., where a(1), a(2), a(3), ..., etc. are coefficients that need to be estimated.

A2: Calculating the transaction volume of each object recognition model sample object in each fourth time period (such as the daily sales volume at the date t), and using the feature data of the third time period corresponding to each object recognition model sample object ( For example, the feature data such as browsing, collection, purchase, transaction, comment, search, etc. of the date t-1 are associated to obtain a plurality of associated data; and based on the obtained associated data, the established initial object recognition model is trained to obtain The actual values of the coefficients a(1), a(2), a(3), etc., to obtain the desired object recognition model.

It should be noted that, in the present embodiment, the regression model can better cross the features, improve the predictive ability, and prevent the merchant from cheating, thereby improving the accuracy of the prediction. The initial object recognition model and the object recognition model may generally be regression models such as a Gradient Boost Regression Tree model. Of course, if the accuracy of the prediction is relatively low, the initial object recognition model and the object recognition model may also adopt a linear model, which is not limited herein.

In addition, it should be noted that after the object recognition model is trained, the object recognition model may be updated in real time or periodically according to the latest object recognition model sample data to improve the accuracy of the object recognition model. In addition, taking a channel object as an example, waiting for the channel After the line is online, the use feature data of each object recognition model sample object in each third time period, and the object use heat of each object recognition model sample object in each corresponding fourth time period may be replaced with each object recognition. The corresponding usage feature data of the model sample object under the channel and the object usage heat and the like are used to better predict the object usage heat of the product object under the channel, and details are not described herein again.

It should be noted that, in this embodiment, steps 201 and 202 are steps of establishing an object recognition model in advance, and are not required to be performed every time the product object is selected, unless the object recognition model trains sample data. A corresponding update has occurred. That is, after performing step 201 and step 202, the subsequent steps may be repeatedly executed multiple times, and details are not described herein.

Step 203: The user terminal receives the product object keyword input by the user, and sends the product object keyword to the server.

Optionally, after receiving the product object keyword input by the user, the user terminal may further extend the received product object keyword to the server, and may expand the received product object keyword and The expanded product object subject is sent to the server to increase the richness of the product object keywords. In addition, the user terminal expands the product object keyword input by the user, and can also avoid the situation that the server needs to simultaneously expand the received large number of product object keywords when a large number of user terminals simultaneously send the product object keyword to the server. Occurs to save the processing resources of the server and reduce the working pressure of the server, thereby further improving the speed and efficiency of subsequent product object selection.

Optionally, the user terminal may augment the received product object keyword in the following manner:

And determining, according to the set sample corpus, a similarity threshold that is not lower than a set similarity threshold (the threshold may be flexibly set according to actual conditions) based on the set sample corpus. At least one sample word; each determined sample word is used as the final desired commodity object subject.

Wherein, the set sample corpus may be a corpus of e-commerce news and the like crawled from an external website by a crawler; and, in addition, based on the set sample corpus, determine a phase with each product object keyword When the at least one sample word is not lower than the set similarity threshold, the language model capable of characterizing the word as a real-value vector, such as the word2vec model, may be firstly trained based on the set sample corpus, and based on the trained word2vec A language model such as a model converts each product object keyword input by the user and each word in the sample corpus into a vector; after that, the calculated similarity calculation formula, such as the Cosine formula, can be used to calculate each word in the sample corpus. The similarity between the keyword objects of each product object input by the user; finally, the word above the value (which may include the value) is selected as the final desired topic word by setting a corresponding similarity threshold.

For example, suppose the user inputs the following three product object keywords to the user terminal according to actual needs: “fast fashion”, “men's clothing”, “women's clothing”, the user terminal may perform the three words based on the set sample corpus. Expansion, such as expansion to get "fashion", "trend", "shirt", "suit", "dress", "jeans" and other words, and the expanded words as the final product object keywords.

Step 204: The server receives the commodity object keyword sent by the user terminal, and acquires a primary selection object that matches the commodity object keyword.

Optionally, the server may search for the corresponding product object title from the product object information of each e-commerce website according to the product object keyword transmitted by the user terminal, and match the product object keyword sent by the user terminal. A product object (such as partial matching, etc.), and each of the searched product objects is a preliminary selection object that matches the product object keyword transmitted by the user terminal.

Each of the product object information in the e-commerce website may include basic information such as an ID (identification), a name (ie, a title), a place of origin, a seller user information, and a category of the product object, and details are not described herein.

In addition, the server may further augment the received product object keyword before acquiring the primary object that matches the product object keyword according to the product object keyword sent by the received user terminal. In order to obtain the corresponding primary selection object based on the expanded commodity object subject terms.

The specific implementation manner in which the server augments the received product object keyword is similar to the specific implementation manner in which the user terminal expands the received product object keyword in step 203, and details are not described herein.

It should be noted that the product object theme input to the user is executed by the server instead of the user terminal. The operation of the word expansion can reduce the performance requirements of the user terminal, so that the method described in the embodiment of the present application has a wider scope of application; in addition, for the user terminal, the processing resources of the user terminal can be saved, and the user terminal can be saved. Work pressure.

Step 205: The server determines, according to the object recognition model obtained by the training and the use feature data of each primary selection object in the first time period, the object usage heat of each primary selection object in the second time period.

For example, assume that the object recognition model is that the server recognizes the volume of the model sample object on each date according to each object, and the browsing, collection, and purchase of each object recognition model sample object on one or more dates before the corresponding date. The object recognition model trained by the feature data such as transaction, comment, search, etc. may be based on the object recognition model, and browse, collect, purchase, and trade according to one or more dates of each primary object before the date t+1. Characteristic data such as comments, searches, etc., predict the volume of each primary object at the date t+1.

It should be noted that, in particular, in this embodiment, when the object usage heat of each commodity object is the turnover amount, since the range of the transaction amount is large, the prediction may not be well predicted, and therefore, the transaction amount may not be directly predicted. First, based on the object recognition model related to the volume, the volume of each commodity object is predicted, and then the transaction amount is multiplied by the corresponding price to improve the accuracy of the prediction.

Step 206: The server selects at least one commodity object from the primary selection objects according to the object usage heat of each primary selection object in the second time period, and matches the keyword of the commodity object in the second time period. Product object.

Optionally, in order to improve the accuracy of the selected commodity object, the server may select at least one object from the primary selection object according to the heat usage of the objects in the second time period of each primary selection object, and the usage heat is not lower than the setting. The product object of the heat threshold (which can be flexibly set according to actual conditions) is used as the product object in the second time period that matches the keyword of the product object.

In addition, the server may also sort the primary objects according to the order in which the object usage heat is used, and take the pre-K (K is any positive integer) primary selection objects as the second time period. A product object whose product object keywords match.

Further, in order to improve the accuracy of the selected commodity object, select from the primary selection object to If one less commodity object is used as the commodity object in the second time period that matches the keyword of the product object, the primary screening object may be manually filtered according to actual needs, or the short-term object is not used hotly. Or a product object whose price does not meet the user's needs (for example, a commodity object with only 5 transactions within three days and a price between 10 and 200 yuan), so as to select a desired commodity object based on the selected primary objects; and / or,

After selecting at least one commodity object from the primary selection object as the commodity object in the second time period that matches the keyword of the commodity object, the selected commodity object may be manually screened according to actual needs, or Deleting the short-term object using the commodity object whose heat is not high or the price does not meet the user's demand, and using the filtered commodity object as the final desired commodity object in the second time period and matching the commodity object keyword I will not repeat them.

Further, since some of the commodity objects have obvious seasonality, such as wearing a dress in June, wearing a coat in September, and the like, the buyer usually purchases the product object in advance, but the product object is in the foregoing. The use of objects in the pre-purchase time period is not very hot, which makes it impossible to place these commodity objects in the front position by the previous prediction method. Therefore, in order to solve such a problem, after determining the heat usage of the objects of the first selected objects in the second time period, the object usage heat of each primary selection object may be adjusted according to the time information, thereby adjusting the order of each primary selection object. In order to select the final desired product object according to the heat-adjusted primary objects. The time series model can be used to predict the heat of the category in the second time period, so that some seasonal commodity objects can emerge in advance to further improve the accuracy of the selection of the commodity object.

That is, before selecting at least one commodity object from the primary selection object as the commodity object in the second time period that matches the commodity object keyword, the method may further include:

Determining each primary selection object according to the set category identification model and the category corresponding to each primary selection object in the category usage heat of one or more historical synchronization time periods corresponding to the second time period The category uses heat in the category of the second time period; wherein the category identification model is trained to represent the category of the commodity object category in the second time period using the heat, and the commodity object category Between the heat usage of the category of one or more historical time periods corresponding to the second time period Identification model of association relationship;

Optionally, for each primary selection object, the category corresponding to the primary selection object may be in the category of the second time period and the category of the primary selected object in the second time period. Using the product of the heat, or a weighted sum of the two (the weights corresponding to the two can be flexibly set according to actual conditions), the updated object in the second time period as the primary selection object uses the heat.

The heat usage of the object is similar to the heat usage of the object, and the heat usage of the category may include at least one or more of a volume, a turnover, and a transaction conversion rate, and the category of the commodity object such as each sample object is used. It can be calculated based on the use characteristic data of the product object such as each sample object, and is not limited thereto. In addition, the historical time period of each time period refers to a historical time period that is on the same calendar day or lunar day as the time period and corresponds to the time period; for example, for the time period January 01, 2016 ~ For the period of January 5, 2016, the historical period of the period can be from January 01, 2015 to January 05, 2015, January 01, 2014 to January 05, 2014, etc. This is not to be repeated.

Optionally, in this embodiment, the category based on the set category identification model and the category corresponding to each primary selection object is one or more historical synchronization time periods corresponding to the second time period. Using the heat, before determining the category corresponding to each primary object in the second time period, the server can obtain the category identification model by:

According to the basic feature data of the sample objects of the various types of mesh recognition models, the pre-established use of the category for predicting the commodity object category in the second time period, and the commodity object category and the second time period The category of one or more historical time periods corresponding to each other is trained using an initial category recognition model of the relationship between the heats to obtain the category identification model.

It should be noted that the fifth time period may be a historical time period; and, similar to the foregoing descriptions regarding the first, second, third, and fourth time periods, the size of the fifth time period may also be based on The actual situation is flexible (although the size of the fifth time period is usually the same as the second time period), which is not limited.

Optionally, taking the heat usage as the volume of the category of the model object of each type of recognition model, the corresponding category identification model can be obtained by the following steps:

B1: Establish an initial category recognition model related to volume.

For example, suppose that each fifth time period can be set to 1 month; the initial category recognition model is a linear model; and it is assumed that the volume of a certain category in this year is z(t), and the volume of the category in the same period last year. For z(t-1), the transaction volume of the category in the same period of the previous year is z(t-2), and so on; then the initial category identification model established can be expressed as z(t)=b(1)*z (t-1)+b(2)*z(t-2)+..., where b(1), b(2), ..., etc. are coefficients that need to be estimated.

It should be noted that the linear model is used as the category recognition model because the model parameters are small and the historical data of the category is relatively stable. Of course, other models, such as a regression model, may be used as the category recognition model to improve the accuracy of the heat prediction using the category, which is not limited herein.

B2: Use the historical data of the category of the latest period of time (such as the latest 3 months of the category history volume of this year) and its corresponding historical data (as in the same period of at least two years on a calendar or lunar calendar) Historical data), the established initial category recognition model is trained to obtain the actual values of the coefficients b(1), b(2), etc., to obtain the final category recognition model.

In addition, it should be noted that after the training of the category identification model is obtained, the category identification model may be updated in real time or periodically according to the latest category identification model sample data to improve the category recognition model. Accuracy; again, taking a product object as an example, after the channel is online, it can also use the heat of the categories of the various target recognition model sample objects in each fifth time period, and various types of object recognition. The model sample object is replaced with a category of the target recognition model sample object under the channel in one or more historical synchronization time periods corresponding to each fifth time period. The corresponding categories use heat, etc., to better predict the heat usage of the category of the product object category under the channel, and will not be described here.

Further, after obtaining the category identification model according to the foregoing manner, if the second time period is determined to be the next month, the category identification model obtained by the training and the class corresponding to each primary selection object may be Determining the category corresponding to each primary object by using the heat of the category in the one or more historical time periods corresponding to the next month (for example, the historical time period of the previous year or the previous two years) The heat is used in the category of the next month.

In addition, it should be noted that, besides predicting the use of the category of the relevant commodity object category for the next month in the above manner, the corresponding initial category recognition model may be established in units of days, such as establishing the following initial class. The recognition model z1(t)=b1(1)*z1(t-1)+b1(2)*z1(t-2)+..., where z1(t) is a category of this year date t The heat used for the purpose, z1(t-1) is the heat used for the category of the same category last year, z1(t-2) is the heat used for the category of the same period of the previous year, and so on; b1(1), b1 (2) Wait for the coefficient to be estimated; after that, use the historical data of the latest period of time (such as 90 days, etc.) and its corresponding historical data to train the established initial category recognition model to obtain b1. The values of the coefficients (1) and b1(2) are used to obtain the final category identification model. After that, the category identification model obtained by the training is used to calculate the 30-day category of each category. The heat, then, the average is used to obtain a more stable category of heat usage for each month, and will not be repeated here.

Further, in order to make the selected commodity object more in line with the user's needs, in order to improve the transaction performance of the commodity object, when the second time period is determined to be a specific time period such as a holiday, the second time period may be corresponding to The various categories of destinations associated with a particular time period are additionally weighted using heat (the degree of additional weighting can be based on actual demand) to ensure that commodity objects corresponding to these categories can emerge in time. For example, the Mid-Autumn Festival moon cake will be hot, and thus, when the second time period is the Mid-Autumn Festival time period, the weight category corresponding to the moon cake can be additionally weighted.

That is, for any of the primary selection objects, the heat is used in the category according to the category corresponding to the primary selection object in the second time period, and the object of the primary selection object in the second time period is Before using the heat to update, the method may further include:

If it is determined that the second time period is a specific time period (such as a Mid-Autumn Festival, a Dragon Boat Festival, and the like, a holiday time period, etc.), if it is determined that the category corresponding to the primary selection object is a specific time corresponding to the second time period The category matched by the segment is based on the set coefficient (the coefficient can be flexibly adjusted according to the actual situation. For example, if the degree of matching between the category and the specific time period is higher, the coefficient can be larger, if the matching degree is lower, Then, the coefficient may be smaller, etc., and the category corresponding to the primary selection object is used to increase the heat usage of the category in the second time period.

Further, since the direct sorting of each commodity object according to the heat usage of the object may result in the appearance of some homogeneous commodity objects, such as "Mid-Autumn Festival Gifts Italian Imports Ferrero Chocolate Roses DIY Gift Boxes Birthday Lovers" and "SF" Italian Ferrero chocolate DIY heart-shaped rose gift box Mid-Autumn Festival birthday gift" These two goods are very similar, if directly displayed to the front desk will lead to a single object. Therefore, before the at least one commodity object is selected from the primary selection objects as the commodity objects in the second time period that match the commodity object subject words, the detachment operations may be performed on the respective primary selection objects.

That is, before the at least one product object is selected from the primary selection object as the commodity object in the second time period that matches the product object keyword, the method may further include:

For each group of at least one primary selection object whose similarity between each other is not lower than a set similarity threshold (the similarity threshold may be the same or different from the first similarity threshold mentioned in the foregoing) a collection of objects, retaining an object in the collection of objects, and deleting other objects, so that at least one commodity object can be selected from each primary selection object obtained after performing the deletion operation as the second time period, The commodity object whose product object keywords match.

That is, the similarity between the respective product objects can be obtained by calculating the similarity of the product object titles. In addition, the similarity calculation can use the JCard similarity formula J(A, B)=|A intersection B|/|A and B| (ie, the similarity of the two titles is the number of common words of the two titles divided by The number of all words in both titles). For example, suppose the title A is "dice chocolate" and the title B is "milk chocolate", then both The similarity is 1/3 because the title intersection has a word "chocolate" and the union of the two titles has three words. Of course, other similar similarity calculation formulas may be used to calculate the similarity between the commodity objects, which is not limited thereto.

Furthermore, for each group of objects consisting of at least one primary selection object whose similarity between them is not lower than a set similarity threshold, when an object in the object set is retained, it is usually retained. The corresponding object uses the object with the highest heat to improve the transaction performance of the commodity object and improve the user's application experience.

In addition, taking a product object as an example for a certain channel, if the channel needs to implement daily update of the product object, the related logic can also be configured so that the product object finally displayed by the channel is not repeated within a few days, here is not Let me repeat.

Further, in this embodiment, in addition to selecting at least one product object from the primary selection object as the commodity object in the second time period that matches the product object keyword, each primary selection object may be performed. In addition to the same operation, the selected ones may be selected in a similar manner after selecting at least one commodity object from the primary selection objects as the commodity objects in the second time period that match the product object keywords. The commodity object performs the same operation.

In addition, the server may select at least one product object from the primary selection object as the product object in the second time period that matches the product object keyword, and may further display the product object information of the selected product object ( The unique identification information such as the ID of the product object or the link of the product object is stored and/or transmitted to the user terminal.

Step 207: The user terminal receives the product object information of the product object that is returned by the server and matches the product object keyword, and uses the product object corresponding to the product object information as the second time period, and the The item object whose product object keyword matches.

Optionally, the user terminal may display the received product object information, and/or display the product object corresponding to the product object information for the user to view, which is not described herein.

In addition, it should be noted that the foregoing step 201, step 203, and step 207 independently constitute a product object selection process executed on the user terminal side, and step 202, step 204 to step 206 are independent. The structure of the product object selection executed on the server side is not described herein.

Further, as shown in FIG. 3 and FIG. 4, the embodiment of the present application further provides two methods for determining a model. Specifically, as shown in FIG. 3, which is a possible flowchart of a method for determining a model in Embodiment 1 of the present application, the method for determining a model may include the following steps:

Step 301: Receive object recognition model training sample data sent by the user terminal.

Step 302: According to the basic feature data of each object recognition model sample object included in the object recognition model training sample data, the pre-established initial object recognition model is trained to obtain a desired object recognition model.

The initial object recognition model is configured to predict a relationship between the use feature data of the product object in the first time period and the object use heat of the product object in the second time period; the first time period is The previous specified time period of the second time period.

The object recognition model is used to characterize the relationship between the use feature data of the commodity object in the first time period and the object use heat of the commodity object in the second time period.

Further, as shown in FIG. 4, which is a possible flow diagram of another method for determining a model in the first embodiment of the present application, the another method for determining a model may include the following steps:

Step 401: Receive the category identification model training sample data sent by the user terminal.

Wherein, the category identification model training sample data includes basic feature data of various target recognition model sample objects, and the basic feature data of each category recognition model sample object includes the category identification model sample object corresponding to the category identification data. The category is in each set sample time period (such as the fifth time period in the foregoing product object selection method; in addition, if the aforementioned product object selection method is not considered, the set sample time period may also be expressed as the first time Segment, in the second time period mentioned later with this model determination method The category usage heat, which is not described in detail, and the category of the category identification model sample object corresponding to one or more historical synchronization time periods corresponding to each set sample time period Use heat.

Step 402: Train the basic feature data of the sample objects of the various types of mesh recognition models included in the sample data according to the category identification model, and train the pre-established initial category recognition model to obtain a desired category recognition model.

The initial category identification model is configured to predict a category usage heat of the commodity object category in the second time period, and one or more historical synchronization time periods corresponding to the commodity object category corresponding to the second time period. The category uses the relationship between the heats.

The category identification model is used to characterize the category usage heat of the commodity object category in the second time period, and the category of one or more historical synchronization time periods corresponding to the commodity object category corresponding to the second time period. Use the relationship between the heats.

In addition, it should be noted that the execution bodies of the model determination method shown in FIG. 3 and FIG. 4 may all be servers; and the specific implementation of each step of the model determination method shown in FIG. 3 and FIG. I will not repeat them.

Further, as shown in FIG. 5, the embodiment of the present application further provides a method for determining the heat usage. The usage heat determination method may include the following steps:

Step 501: Acquire usage characteristic data of each commodity object in a first time period;

Step 502: based on the relationship between the use feature data of the product object in the first time period and the object use heat of the product object in the second time period, and the use feature data of each commodity object in the first time period, Determining the object usage heat of each commodity object in the second time period;

The first time period is a previous specified time period of the second time period; and the association relationship is based on the use feature data of each sample object in each third time period and each corresponding sample object in each corresponding The object of the four time period is established using the heat; the third time period is the previous specified time period of the corresponding fourth time period.

It should be noted that the association relationship is similar to the set object recognition model described above. In addition, the specific implementation of each step of using the heat determination method shown in FIG. 5 can be referred to the foregoing related description, and no further description is provided herein. .

According to the content of the first embodiment of the present application, in the solution according to the first embodiment of the present application, at least one product object can be automatically selected from the mass object based on the product object keyword input by the user and the set object recognition model. The ultimate object that meets the user's needs, thereby greatly improving the efficiency of the selection of commodity objects, thereby reducing the cost of manual inventory and improving operational efficiency.

In addition, since the object can be used in the second time period according to each of the primary objects, at least one object selected from the primary objects is used as the final desired product object, and the product object having the heat not lower than the set heat threshold is used as the final desired product object. Can improve the accuracy of the selection of commodity objects.

In addition, since the time series model, that is, the category identification model, can be used to calculate the category of the category, the heat is used, and the order of each commodity object is adjusted according to the category, so that the season can be Holidays and other timely adjustment of commodity objects to further reduce the cost of manual inventory and improve operational efficiency.

Finally, it should be noted that the solution described in the embodiment of the present application has no language, software or hardware limitation, and can be implemented based on a general cloud computing platform. However, in order to improve the efficiency of the selection of the channel object, it is preferred to use a high-performance programming language (such as C, C++, or Java) and high-performance hardware, etc., which is not described in this embodiment.

Embodiment 2:

Based on the same inventive concept as the first embodiment of the present application, the second embodiment of the present application provides a product object selection device. For the specific implementation of the product object selection device, refer to the related description of the server in the first embodiment of the method. The details of the product object selection device may include:

The keyword receiving unit 601 is configured to receive a product object keyword sent by the user terminal;

The object obtaining unit 602 is configured to obtain a primary selection object that matches the keyword of the product object;

The heat determining unit 603 can be used to identify the model based on the object and each of the primary objects in the first Using the feature data for a period of time, determining the object usage heat of each of the preliminary objects in the second time period; wherein the object recognition model is the trained use characteristic data for characterizing the commodity object in the first time period And a recognition model of an association relationship between the object usage heat of the commodity object in the second time period; the first time period being a previous specified time period of the second time period;

The object screening unit 604 is configured to select, according to the object usage heat of the first selected object in the second time period, at least one commodity object from the primary selection object as the keyword in the second time period Matching product objects.

Optionally, the object screening unit 604 is specifically configured to select at least one object from the primary selection object to use a commodity object whose heat is not lower than a set heat threshold, as the second time period, and the commodity object theme. The product object that matches the word.

Optionally, the commodity object selection device may further include an object recognition sample data receiving unit 605 and an object recognition model determining unit 606:

The object identification sample data receiving unit 605 is configured to determine an object of each primary selection object in the second time period based on the set object recognition model and the usage feature data of each primary selection object in the first time period. Before using the heat, receiving the object recognition model training sample data sent by the user terminal, wherein the object recognition model training sample data includes basic feature data of each object recognition model sample object, and each object identifies a model sample object basis The feature data includes usage feature data of the object recognition model sample object in each third time period, and object usage heat of the object recognition model sample object in each corresponding fourth time period; the third time period is a corresponding specified time period of the corresponding fourth time period;

The object recognition model determining unit 606 is configured to: according to the basic feature data of each object recognition model sample object, use the pre-established use feature data for predicting the commodity object in the first time period, and the commodity object in the second time The object of the segment is trained using an initial object recognition model of the relationship between the heats to obtain the object recognition model.

Optionally, the commodity object selection device may further include a category heat determination unit 607 and an object heat update unit 608:

The category heat determining unit 607 is configured to: before selecting at least one product object from the primary selection object as the commodity object in the second time period that matches the product object keyword, based on the set class Determining the model, and the category corresponding to each primary object in the category of one or more historical time periods corresponding to the second time period, determining the category corresponding to each primary object in the The category of the second time period uses heat; wherein the category identification model is trained to represent the category of the commodity object category in the second time period, and the commodity object category is in the second A recognition model of the relationship between the heat usage of the category of one or more historical time periods corresponding to the time period;

The object heat update unit 608 may be configured to use, for each primary selection object, a heat in the category of the second time period according to the category corresponding to the primary selection object, and the second selected time in the second time period The objects of the segment are updated with heat.

Optionally, the commodity object selection device may further include a category identification sample data receiving unit 609 and a category identification model determining unit 610:

The category identification sample data receiving unit 609 is configured to: at the one or more historical synchronization times corresponding to the second time period, based on the set category identification model and the category corresponding to each primary selection object The category of the segment uses the heat to determine that the category corresponding to each of the primary selected objects receives the category identification model training sample data sent by the user terminal before the category usage heat of the second time period, wherein the category identification The model training sample data includes basic feature data of various target recognition model sample objects, and the basic feature data of each category recognition model sample object includes the category corresponding to the category identification model sample object at each fifth time The category usage heat of the segment, and the category corresponding to the category identification model sample object uses the heat in the category of one or more historical synchronization time periods corresponding to each fifth time period;

The category identification model determining unit 610 is configured to use, according to the basic feature data of the sample objects of the various types of mesh recognition models, the heat usage and the commodity object used for predicting the category of the commodity object category in the second time period. The category is trained in an initial category recognition model of the association relationship between the categories of heat usage of one or more historical time periods corresponding to the second time period, and the category is obtained Identify the model.

Optionally, the commodity object selection device may further include a category heat update unit 611:

The category heat update unit 611 may be configured to use the heat in the category of the second time period according to the category corresponding to the primary selection object for any primary selection object, and the primary selection object is in the Before determining that the second time period is a specific time period, the object corresponding to the first time period is determined to be a specific time corresponding to the second time period. The category matched by the segment increases the category usage heat of the category corresponding to the primary selection object in the second time period according to the set coefficient.

Optionally, the commodity object selection device may further include a commodity object de-same unit 612:

The commodity object de-same unit 612 can be configured to select at least one commodity object from the primary selection object as the commodity object in the second time period that matches the product object keyword, according to each primary selection object. Correlating degrees between corresponding titles, determining similarity between each primary selection object; for each set of object collections consisting of at least one primary selection object whose similarity between the two is not lower than the first similarity threshold , retain an object in the collection of objects, and delete other objects.

Optionally, the commodity object selection device may further include a keyword expansion unit 613:

The keyword expansion unit 613 may be configured to determine, for each product object keyword sent by the user terminal, based on the set sample corpus, before acquiring the primary object that matches the product object keyword The similarity between the subject keywords is not lower than the at least one sample word of the second similarity threshold; the determined sample words are taken as the final desired commodity object keywords.

In addition, it should be noted that the object recognition model may be a regression model; the category recognition model may be a linear model.

Further, based on the same inventive concept as the first embodiment of the present application, the second embodiment of the present application further provides another product object selection device. For the specific implementation of the other product object selection device, refer to the first embodiment of the method. The relevant description of the user terminal, the repetition will not be repeated, as shown in Figure 7. As shown, the other commodity object selection device can mainly include:

The keyword receiving unit 701 is configured to receive a product object keyword input by the user;

a keyword sending unit 702, configured to send the product object keyword to a server;

The object information receiving unit 703 is configured to receive the commodity object information returned by the server according to the product object keyword;

The object determining unit 704 is configured to determine the product object corresponding to the product object information as the product object that matches the product object keyword in the second time period;

Optionally, the another commodity object selection device may further include an object recognition sample data receiving unit 705 and an object recognition sample data transmitting unit 706:

The object recognition sample data receiving unit 705 is configured to receive the object recognition model training sample data input by the user before receiving the commodity object keyword input by the user, wherein the object recognition model training sample data includes each object recognition model Base feature data of the sample object, and the base feature data of each object recognition model sample object includes usage feature data of the object recognition model sample object in each third time period, and the object recognition model sample object is in each The object of the corresponding fourth time period uses the heat; the third time period is the previous specified time period of the corresponding fourth time period;

The object identification sample data sending unit 706 is configured to send the object recognition model training sample data to a server, and the server identifies the basic features of each object recognition model sample object included in the sample data according to the object recognition model. Data, used in advance for predicting the use of feature data of the product object in the first time period, and the use of the object object in the second time period The initial object recognition model of the relationship between the two is trained to obtain the set object recognition model.

Optionally, the another commodity object selection device may further include a category identification sample data receiving unit 707 and a category identification sample data sending unit 708:

The category identification sample data receiving unit 707 is configured to receive, after receiving the product object information returned by the server according to the product object subject word, the category identification model training sample data input by the user, wherein the category The recognition model training sample data includes basic feature data of various target recognition model sample objects, and the basic feature data of each category recognition model sample object includes the category corresponding to the category identification model sample object in each fifth The category usage heat of the time period, and the category corresponding to the category identification model sample object uses the heat in the category of one or more historical synchronization time periods corresponding to each fifth time period;

The category identification sample data sending unit 708 is configured to send the category identification model training sample data to a server, and the server trains the sample of various types of eye recognition models included in the sample data according to the category identification model. The basic feature data of the object, for the pre-established category use heat for predicting the item object category in the second time period, and one or more historical time periods corresponding to the item object category corresponding to the second time period The category of the segment is trained using the initial category recognition model of the relationship between the heats, and the category used to characterize the commodity object category in the second time period is used, and the commodity object category is in the second time. The category identification model of the association relationship between the heats of the one or more historical time periods corresponding to the segments.

Further, based on the same inventive concept as the first embodiment of the present application, the second embodiment of the present application further provides a model determining device. For the specific implementation of the model determining device, refer to the first model in the first embodiment of the method. For the description of the method, the description of the method will not be repeated. As shown in FIG. 8, the model determining device may mainly include:

The data receiving unit 801 is configured to receive the object recognition model training sample data sent by the user terminal, where the object recognition model training sample data includes basic feature data of each object recognition model sample object, and each object recognition model sample The base feature data of the object includes the object Identifying the use feature data of the model sample object in each of the third time periods, and the object use heat of the object recognition model sample object in each corresponding fourth time period; the third time period is the corresponding fourth time period The previous specified time period;

The model training unit 802 is configured to: according to the basic feature data of each object recognition model sample object included in the object recognition model training sample data, the pre-established use feature data for predicting the commodity object in the first time period, Training with the initial object recognition model of the relationship between the object objects in the second time period and the heat usage of the objects, obtaining the use feature data for characterizing the product object in the first time period, and the second time period with the product object The object uses an object recognition model of the relationship between the heats; the first time period is a previous specified time period of the second time period.

Further, based on the same inventive concept as the first embodiment of the present application, the second embodiment of the present application further provides another model determining device. For the specific implementation of the other model determining device, refer to the related method in the first embodiment of the method. Another description of the method for determining the model is not repeated here. As shown in FIG. 9, the other model determining device may mainly include:

The data receiving unit 901 is configured to receive the category identification model training sample data sent by the user terminal, where the category identification model training sample data includes basic feature data of various target recognition model sample objects, and each class The base feature data of the target model sample object includes the category corresponding to the category object of the sample identification model, and the fifth time period in the sample object selection method (in the foregoing, the fifth time period in the foregoing commodity object selection method; The commodity object selection method may also represent the set sample time period as the first time period to distinguish the second time period mentioned later by the model determining device, and the category use heat is not described herein. And the category corresponding to the category identification model sample object uses heat in a category of one or more historical synchronization time periods corresponding to each set sample period;

The model training unit 902 is configured to: according to the category identification data of the category identification model sample objects included in the sample data of the category identification model, the pre-established class for predicting the commodity object category in the second time period The initial category identification of the relationship between the heat usage and the heat usage of the category of the commodity object category in one or more historical time periods corresponding to the second time period The model is trained to obtain a category used to characterize the category of the commodity object category in the second time period, and the category of the commodity object category in the one or more historical time periods corresponding to the second time period. A category identification model for the relationship between heats.

Further, based on the same inventive concept as the first embodiment of the present application, the second embodiment of the present application further provides a heat determining device, and the specific implementation of the heat determining device can be referred to the heat usage in the first embodiment of the method. For the description of the method, the description of the method will not be repeated. As shown in FIG. 10, the heat determining device may mainly include:

The data obtaining unit 1001 is configured to obtain usage characteristic data of each commodity object in a first time period;

The heat determining unit 1002 is configured to use an association relationship between the use feature data of the product object in the first time period and the object use heat of the product object in the second time period, and the product object in the first time period Using the feature data to determine the object usage heat of each commodity object in the second time period;

Finally, it should be noted that any of the names in the specification of the present application and the drawings (such as the first time period, the second time period, etc.) are used for distinguishing only, without any limiting meaning.

Those skilled in the art will appreciate that embodiments of the present application can be provided as a method, apparatus (device), or computer program product. Thus, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment in combination of software and hardware. Moreover, the application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) including computer usable program code.

The present application is described with reference to flowchart illustrations and/or block diagrams of a method, apparatus, and computer program product according to embodiments of the present application. It should be understood that the flowcharts and/or may be implemented by computer program instructions. Each of the processes and/or blocks in the block diagrams, and combinations of the flows and/or blocks in the flowcharts and/or block diagrams. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing device to produce a machine for the execution of instructions for execution by a processor of a computer or other programmable data processing device. Means for implementing the functions specified in one or more of the flow or in a block or blocks of the flow chart.

The computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture comprising the instruction device. The apparatus implements the functions specified in one or more blocks of a flow or a flow and/or block diagram of the flowchart.

These computer program instructions can also be loaded onto a computer or other programmable data processing device such that a series of operational steps are performed on a computer or other programmable device to produce computer-implemented processing for execution on a computer or other programmable device. The instructions provide steps for implementing the functions specified in one or more of the flow or in a block or blocks of a flow diagram.

While the preferred embodiment of the present application has been described, it will be apparent that those skilled in the art can make further changes and modifications to the embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and the modifications and

It will be apparent to those skilled in the art that various modifications and changes can be made in the present application without departing from the spirit and scope of the application. Thus, it is intended that the present invention cover the modifications and variations of the present invention.

Claims

A method for selecting a commodity object, comprising:

Receiving a commodity object keyword sent by the user terminal;

Obtaining a primary selection object that matches the subject object of the commodity object;

Determining the object usage heat of each primary object in the second time period based on the set object recognition model and the use feature data of each of the preliminary objects in the first time period; wherein the object recognition model is used for training And a recognition model for characterizing the relationship between the use feature data of the commodity object in the first time period and the object use heat of the commodity object in the second time period; the first time period is the previous one of the second time period Specified time period;

And selecting at least one commodity object from the primary selection objects as the commodity object matching the commodity object keyword in the second time period according to the object usage heat of each of the primary selection objects in the second time period.
The method of claim 1, wherein the selecting at least one item from the preliminary objects as the item of the item in the second time period that matches the subject of the item comprises:

Selecting at least one object from the primary selection object uses a commodity object having a heat not lower than a set heat threshold as a commodity object in the second time period that matches the commodity object keyword.
The method according to claim 1, wherein the object usage of each of the primary objects in the second time period is determined based on the set object recognition model and the usage characteristic data of each of the preliminary objects in the first time period. Before the heat, the method further includes:

Receiving object recognition model training sample data sent by the user terminal, wherein the object recognition model training sample data includes basic feature data of each object recognition model sample object, and the basic feature data of each object recognition model sample object includes Using the feature data of the object recognition model sample object in each third time period, and the object use heat of the object recognition model sample object in each corresponding fourth time period; the third time period is the corresponding fourth The previous specified time period of the time period;

Determining, according to the basic feature data of each object recognition model sample object, a pre-established relationship between the use feature data for predicting the product object in the first time period and the object use heat of the product object in the second time period The initial object recognition model is trained to obtain the object recognition model.
The method of claim 1 wherein said method selects at least one item of merchandise from said preliminary object as said item of merchandise in said second period of time that matches said item subject matter Also includes:

Determining each primary selection object according to the set category identification model and the category corresponding to each primary selection object in the category usage heat of one or more historical synchronization time periods corresponding to the second time period The category uses heat in the category of the second time period; wherein the category identification model is trained to represent the category of the commodity object category in the second time period using the heat, and the commodity object category An identification model of an association relationship between the heat usages of the one or more historical synchronization time periods corresponding to the second time period;

For each primary selection object, the hotspot of the second time period is used to update the object usage heat in the second time period according to the category corresponding to the primary selection object.
The method of claim 4, wherein the category identification model based on the set, and the category corresponding to each of the preliminary selected objects are in one or more historical synchronization times corresponding to the second time period The category of the segment uses the heat to determine that the category corresponding to each primary object is before the category usage heat of the second time period, and the method further includes:

Receiving the category identification model training sample data sent by the user terminal, wherein the category identification model training sample data includes basic feature data of various target recognition model sample objects, and each category identifies the model sample object basis The feature data includes the category usage heat of the category corresponding to the category identification model sample object in each fifth time period, and the category corresponding to the category identification model sample object corresponds to each fifth time period. The use of heat in one or more historical time periods;

According to the basic feature data of the sample objects of the various types of mesh recognition models, the pre-established use of the category for predicting the commodity object category in the second time period is used, and the commodity object category corresponds to the second time period. The initial relationship between the heat usage of the category of one or more historical time periods The category recognition model is trained to obtain the category recognition model.
The method according to claim 4 or 5, wherein the historical time period of each time period refers to a historical time period that is on the same calendar day or lunar day as the time period and corresponds to the time period. .
The method according to claim 4, wherein for any of the preliminary objects, the primary selected object is used in a category according to the category corresponding to the primary selected object in the second time period Before the object of the second time period is updated by using the heat, the method further includes:

If it is determined that the second time period is a specific time period, if it is determined that the category corresponding to the primary selection object is a category that matches a specific time period corresponding to the second time period, according to the set coefficient And increasing the category usage heat of the category corresponding to the primary selection object in the second time period.
The method of claim 1 wherein said method selects at least one item of merchandise from said preliminary object as said item of merchandise in said second period of time that matches said item subject matter Also includes:

Determining the similarity between the primary selected objects according to the similarity between the titles corresponding to the primary objects;

For each set of objects consisting of at least one primary selected object whose degree of similarity is not lower than the first similarity threshold, one object in the set of objects is retained, and other objects are deleted.
The method of claim 1, wherein the method further comprises: prior to obtaining the primary selected object that matches the subject matter of the product object, the method further comprising:

Determining, according to the set sample corpus, at least one sample word having a similarity between the product object subject and the second similarity threshold based on the set sample corpus;

The determined sample words are used as the final desired commodity object keywords.
The method of claim 4 wherein said object recognition model is a regression model; said category recognition model is a linear model.
A method for selecting a commodity object, comprising:

Receiving a product object keyword input by the user, and transmitting the product object keyword to the service Device

Receiving the product object information returned by the server according to the product object keyword, and using the product object corresponding to the product object information as the product object matching the product object keyword in the second time period;

The product object information is used by the server according to the set object recognition model and the primary selection objects that match the product object topic in the first time period, and the product data from the product. Corresponding information of the commodity object selected in the primary object matched by the object keyword; the first time period is a previous specified time period of the second time period;

The set object recognition model is used to characterize the relationship between the use feature data of the product object in the first time period and the object use heat of the product object in the second time period.
The method of claim 11, wherein the method further comprises: before receiving the item object keyword input by the user, the method further comprising:

Receiving object recognition model training sample data input by the user, wherein the object recognition model training sample data includes basic feature data of each object recognition model sample object, and the basic feature data of each object recognition model sample object includes the The object recognition model sample object uses feature data in each third time period, and the object recognition model sample object uses heat in each corresponding fourth time period; the third time period is a corresponding fourth time The previous specified time period of the segment;

Transmitting the object recognition model training sample data to a server, wherein the server identifies the basic feature data of each object recognition model sample object included in the sample data according to the object recognition model, and the pre-established product object for predicting The initial object recognition model using the relationship between the feature data and the object use heat in the second time period in the first time period is trained to obtain the set object recognition model.
A model determining method, comprising:

Receiving object recognition model training sample data sent by the user terminal, wherein the object recognition model training sample data includes basic feature data of each object recognition model sample object, and the basic feature data of each object recognition model sample object includes Object recognition model sample object in each Using the feature data in the three time periods, and the object recognition heat of the object recognition model sample object in each corresponding fourth time period; the third time period is a previous specified time period of the corresponding fourth time period;

And according to the basic feature data of each object recognition model sample object included in the object recognition model training sample data, the pre-established use feature data for predicting the product object in the first time period, and the commodity object in the second time The object of the segment is trained using the initial object recognition model of the association relationship between the heats, and is obtained between the use feature data for characterizing the commodity object in the first time period and the object use heat of the commodity object in the second time period. An object recognition model of the association relationship; the first time period is a previous specified time period of the second time period.
The method according to claim 13, wherein the usage feature data comprises at least one or more of a number of views, a number of favorites, a number of purchases, a number of transactions, a number of comments, and a number of searches; The object usage heat includes at least one or more of volume, turnover, and transaction conversion rate.
A model determining method, comprising:

Receiving the category identification model training sample data sent by the user terminal, wherein the category identification model training sample data includes basic feature data of various target recognition model sample objects, and each category identifies the model sample object basis The feature data includes a category usage heat of the category corresponding to the category identification model sample object in each first time period, and the category corresponding to the category identification model sample object corresponds to each first time period. The use of heat in one or more historical time periods;

Generating the basic feature data of the sample objects of the various types of mesh recognition models included in the sample data according to the category identification model, and using the pre-established category for predicting the use of the commodity object category in the second time period, and the commodity object The category is trained in an initial category recognition model of the association relationship between the categories of heat usage using one or more historical time periods corresponding to the second time period, and is obtained in the second category for characterizing the commodity object. The category of the time period uses a heat classification, a category recognition model that is related to the heat usage of the category of the commodity object category in one or more historical time periods corresponding to the second time period.
The method of claim 15 wherein said category usage heat comprises at least one or more of a volume, a turnover, and a transaction conversion rate.
A method for determining heat usage, comprising:

Obtaining usage characteristic data of each commodity object in a first time period;

Determining each item based on the relationship between the use feature data of the product object in the first time period and the object use heat of the product object in the second time period, and the use feature data of each product object in the first time period The object uses heat in the second time period;

The first time period is a previous specified time period of the second time period; and the association relationship is based on usage characteristic data of each sample object in each third time period and each sample object is in each corresponding fourth time The object of the segment is established using the heat; the third time period is the previous specified time period of the corresponding fourth time period.
A commodity object selection device, comprising:

a keyword receiving unit, configured to receive a commodity object keyword sent by the user terminal;

An object obtaining unit, configured to acquire a primary selection object that matches the keyword of the commodity object;

a heat determining unit, configured to determine, according to the set object recognition model and the use feature data of each of the preliminary objects in the first time period, the object usage heat of each of the preliminary objects in the second time period; wherein the object recognition The model is a training model for characterizing the relationship between the use feature data of the commodity object in the first time period and the heat usage of the object object in the second time period; the first time period is The previous specified time period of the second time period;

An object screening unit, configured to select at least one commodity object from the primary selection object as the second time period according to the object usage heat of each primary selection object in the second time period, and the subject matter of the commodity object Matching item object.
A commodity object selection device, comprising:

a keyword receiving unit, configured to receive a commodity object keyword input by the user;

a keyword sending unit, configured to send the commodity object keyword to a server;

An object information receiving unit, configured to receive, according to the product object keyword, the server returns Product object information;

An object determining unit, configured to determine a product object corresponding to the product object information as a product object that matches the product object keyword in a second time period;

The product object information is used by the server according to the set object recognition model and the primary selection objects that match the product object topic in the first time period, and the product data from the product. Corresponding information of the commodity object selected in the primary object matched by the object keyword; the first time period is a previous specified time period of the second time period;

The set object recognition model is used to characterize the relationship between the use feature data of the product object in the first time period and the object use heat of the product object in the second time period.
A model determining device, comprising:

a data receiving unit, configured to receive object recognition model training sample data sent by the user terminal, where the object recognition model training sample data includes basic feature data of each object recognition model sample object, and each object recognition model sample object The basic feature data includes the use feature data of the object recognition model sample object in each third time period, and the object use heat of the object recognition model sample object in each corresponding fourth time period; the third time The segment is a previous specified time period of the corresponding fourth time period;

a model training unit, configured to perform, according to the basic feature data of each object recognition model sample object included in the sample identification model training sample data, the pre-established use characteristic data for predicting the commodity object in the first time period, and The commodity object is trained in the initial object recognition model of the relationship between the objects using the heat in the second time period, and the use feature data for characterizing the commodity object in the first time period and the second time period of the commodity object are obtained. An object recognition model in which an object uses an association relationship between heats.
A model determining device, comprising:

a data receiving unit, configured to receive the category identification model training sample data sent by the user terminal, wherein the category identification model training sample data includes basic feature data of various target recognition model sample objects, and each category Identifying the base feature data of the model sample object includes the category recognition module The category corresponding to the sample object uses the heat in the category of each first time period, and the category corresponding to the category identification model sample object is in one or more historical synchronization time periods corresponding to each first time period. Category of use;

a model training unit, configured to: according to the category identification model, the basic feature data of the sample objects of the various types of mesh recognition models included in the sample data, and the pre-established categories for predicting the category of the commodity object in the second time period Training is performed using the initial category recognition model of the relationship between the heat and the category of the commodity object category in the one or more historical time periods corresponding to the second time period, and is used to characterize the commodity. The category of the object category in the second time period is the category of the association between the heat usage and the category usage heat of the commodity object category in one or more historical time periods corresponding to the second time period. Identify the model.
A heat determining device is characterized in that it comprises:

a data obtaining unit, configured to acquire usage characteristic data of each commodity object in a first time period;

a heat determining unit, configured to use an association relationship between the use feature data of the product object in the first time period and the object use heat of the product object in the second time period, and use of each commodity object in the first time period Feature data, determining an object usage heat of each commodity object in a second time period; the first time period is a previous specified time period of the second time period; and the association relationship is based on each sample object in each third The usage characteristic data of the time period and the object usage heat of each sample object in each corresponding fourth time period; the third time period is a previous specified time period of the corresponding fourth time period.