WO2019148982A1

WO2019148982A1 - Sorting

Info

Publication number: WO2019148982A1
Application number: PCT/CN2018/121305
Authority: WO
Inventors: 唐金川
Original assignee: 北京三快在线科技有限公司
Priority date: 2018-01-31
Filing date: 2018-12-14
Publication date: 2019-08-08
Also published as: CN108335137B; CN108335137A

Abstract

Provided are a sorting method and apparatus, an electronic device, and a computer readable medium. The sorting method comprises: obtaining a plurality of objects to be sorted according to request information (S110); obtaining a sorting feature value of each of the objects (S120); and sorting the plurality of objects according to the sorting feature value of each object (S130); wherein the sorting feature value of each of the objects is obtained based on a click feature value, a click weight, a conversion feature value, a conversion weight, a transaction feature value, and a transaction weight of each of the objects.

Description

Sort

Cross-reference to related applications

The present application claims priority to Chinese Patent Application No. 201810097006.3 filed on Jan. 31, 2018, entitled "Sorting Method and Apparatus, Electronic Apparatus, Computer Readable Medium", the entire disclosure of which is incorporated by reference. The way is incorporated in this article.

Technical field

The present disclosure relates to a sorting method and apparatus, an electronic device, and a computer readable medium in the field of Internet technology.

Background technique

The sorting optimization mechanism is an important part of the search and recommendation system, such as the advertising system. For performance-based advertising, the advertising platform can charge for clicks, for example, CPC (Cost Per Click) billing, based on the number of times the ad is clicked (clicks).

Summary of the invention

The present disclosure provides a sorting method and apparatus, an electronic device, and a computer readable medium.

According to an aspect of the present disclosure, there is provided a sorting method comprising: acquiring a plurality of objects to be sorted according to request information; acquiring a sorting feature value of each of the objects; and selecting the sorting feature value according to each of the objects Sorting the plurality of objects; wherein the sorting feature values of each of the objects are obtained based on a click feature value, a click weight, a conversion feature value, a conversion weight, a transaction feature value, and a transaction weight of each of the objects.

In an exemplary embodiment of the present disclosure, acquiring the sorting feature value of the object includes: acquiring a historical exposure amount, a historical click amount, and a historical order quantity of the object; according to the request information and the object The historical exposure amount and the historical click amount obtain a click feature value of the object; and obtain the location according to the request information, the click feature value of the object, the historical click amount, and the historical order quantity a conversion feature value of the object; acquiring, according to the request information and the conversion feature value of the object, a transaction feature value of the object; and the click feature value, the conversion feature value, and the object according to the object The transaction feature value calculates the ranking feature value of the object.

In an exemplary embodiment of the present disclosure, obtaining the click feature value of the object according to the request information and the historical exposure amount and the historical click amount of the object includes: according to the object The historical exposure amount and the historical hit amount acquire a historical exposure click rate of the object; and obtain a click feature value of the object according to the object attribute of the object and the historical exposure click rate and the request information.

In an exemplary embodiment of the present disclosure, obtaining the transformed feature value of the object according to the request information, the click feature value of the object, the historical click amount, and the historical order quantity includes: The historical click amount of the object and the historical order quantity to obtain the historical click order rate of the object; and the object attribute according to the object, the historical click order rate, and the click feature value and the The request information obtains a conversion feature value of the object.

In an exemplary embodiment of the present disclosure, obtaining the transaction feature value of the object according to the request information and the converted feature value of the object includes: obtaining the location according to the object attribute of the object and the request information Determining the transaction amount of the object; normalizing the predicted transaction amount of the object by using a preset reference transaction amount; and calculating the normalized transaction amount according to the object and the object Converting the feature value obtains a transaction feature value for the object.

In an exemplary embodiment of the present disclosure, sorting the plurality of objects according to the sorting feature value of each of the objects includes: determining, based on the click feature values of the plurality of objects a reference click feature value for normalizing the click feature value; determining a reference conversion feature value for normalizing the transformed feature value based on the converted feature value of the plurality of objects; Determining a reference click feature value for normalizing the transaction feature value based on the transaction feature value of the plurality of objects; for each of the objects, using the reference click feature value to the object Normalizing the click feature value, normalizing the converted feature value of the object by using the reference conversion feature value, and using the reference transaction feature value to the transaction feature value of the object Performing normalization, and calculating the object based on the normalized click feature value, the normalized transformed feature value, and the normalized transaction feature value of the object Feature value sequence; and the characteristic value according to each sort of the object, of the plurality of objects to be sorted.

In an exemplary embodiment of the present disclosure, the sorting method further includes: for each of the objects, obtaining a click weight, a conversion weight, and a transaction weight of the object according to a current state of the object.

In an exemplary embodiment of the present disclosure, acquiring the click weight, the conversion weight, and the transaction weight of the object according to the current state of the object includes: setting the location for the object when the object is in the first state The click right is greater than the conversion weight and the conversion weight is greater than the transaction weight; when the object is in the second state, the conversion right is set to be greater than the transaction weight and the transaction right is greater than the transaction weight for the object The click weight; when the object is in the third state, the transaction right is set to be greater than the conversion weight and the conversion weight is greater than the click weight for the object; wherein the click weight of the object The sum of the conversion weight and the transaction weight is a preset constant.

In an exemplary embodiment of the present disclosure, when the object is in the first state, setting the click right to the conversion weight and the conversion right is greater than the transaction weight for the object includes: When the consumption budget ratio of the object is within the first preset range, the click weight of the object is increased according to the consumption budget ratio of the object.

In an exemplary embodiment of the present disclosure, when the object is in the second state, setting the conversion right to be equal to the transaction weight and the transaction right is greater than the click weight for the object includes When the consumption budget ratio of the object is within the second preset range, the conversion weight of the object is increased according to the historical exposure conversion rate of the object.

In an exemplary embodiment of the present disclosure, when the object is in the third state, setting the transaction right is greater than the conversion weight for the object, and the conversion weight is greater than the click weight: when the When the consumption budget of the object is within the third preset range, the transaction weight of the object is increased according to the transaction amount of the object.

In an exemplary embodiment of the present disclosure, the request information includes search information input by a current user; and/or combined information between the current user and each object; and/or a user of the current user Attributes.

According to an aspect of the present disclosure, a sorting apparatus is provided, including: an object obtaining module, configured to acquire a plurality of objects to be sorted according to the request information; and a feature value obtaining module, configured to acquire a sorting feature value of each of the objects; And a sorting module, configured to sort the plurality of objects according to the sorting feature value of each of the objects; wherein a sorting feature value of each of the objects is based on a click feature value of each of the objects, and clicking Weights, conversion eigenvalues, conversion weights, transaction eigenvalues, and transaction weights are obtained.

According to an aspect of the present disclosure, an electronic device is provided, including a memory, a processor, and a computer program stored on the memory and executable on the processor, the program being executed by the processor to implement any of the above embodiments Method steps in .

According to an aspect of the present disclosure, there is provided a computer readable medium having stored thereon a computer program, the program being executed by a processor to implement the method steps of any of the above embodiments.

According to the sorting method and device, the electronic device, and the computer readable medium in some embodiments of the present disclosure, the ranking feature values of each object are obtained based on the transaction feature values of the objects to be sorted and the corresponding transaction weights. Accurate object sorting; on the other hand, it can achieve more accurate delivery.

The above general description and the following detailed description are merely exemplary and are not intended to limit the disclosure.

DRAWINGS

The above and other objects, features, and advantages of the present invention will become more apparent from the aspects of the description.

FIG. 1 is a flow chart showing a sorting method according to an exemplary embodiment.

2 is a flow chart of a sorting method shown in accordance with another exemplary embodiment.

FIG. 3 is a schematic diagram of a sorting device, according to an exemplary embodiment.

FIG. 4 is a schematic diagram of an electronic device, according to an exemplary embodiment.

Detailed ways

Example embodiments will now be described more fully with reference to the accompanying drawings. However, the example embodiments can be embodied in a variety of forms and should not be construed as being limited to the examples set forth herein; rather, these embodiments are provided to make the disclosure more complete and complete, and to fully convey the concept of the example embodiments. Those skilled in the art. The drawings are only schematic representations of the disclosure, and are not necessarily to scale. The same reference numerals in the drawings denote the same or similar parts, and the repeated description thereof may be omitted.

Furthermore, the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, various embodiments are provided to provide a full understanding of the technical solutions of the present disclosure. However, one skilled in the art will appreciate that one or more of the specific details may be omitted or other methods, components, devices, steps, etc. may be employed. In other instances, various aspects of the present disclosure may be obscured without the details of the structure, method, apparatus, implementation, or operation.

When optimizing the sorting mechanism in the advertising system, in addition to optimizing platform revenue, the user experience is also optimized. For platforms that cannot track transactions, search engines such as Google and Baidu can optimize the user experience by optimizing the click rate (the ratio of the number of times a content on a platform is clicked to the number of times displayed). In other words, the user experience of a platform that cannot track transactions is measured by the clickthrough rate. A better user experience means that users can quickly find the page they want and leave the page quickly after clicking. In this case, you can choose to optimize the overall click-through rate of the platform. In this way, when the click rate is increased, not only the user experience is optimized, but also the platform can obtain more advertising revenue. The sorting formula for the sorting mechanism optimized by click rate is:

Sort score = bid * click rate.

The platform that can track transactions, such as Amazon, Taobao and other e-commerce websites, in addition to the click rate measurement in the above sorting mechanism, the user experience is measured to increase the conversion rate to measure the user experience. Conversion rate is the product of the click-through rate and the click-to-order rate (the ratio of the number of orders placed on the platform to the number of times the order was clicked). For the platform that can track the transaction, the user's experience is reflected in “shopping” and “buying”, so when optimizing the user experience, the overall click rate and conversion rate of the platform can be optimized, and the two are weighted by different weights w1 and w2. The form of the sum is combined. The optimization of conversion rate can also bring more conversion orders to advertisers, which will bring more benefits to advertisers to some extent. under these circumstances,

User experience score = w1 * click rate + w2 * conversion rate.

The sorting formula for sorting mechanisms optimized by clickthrough rate and conversion rate is:

Sort score = bid * (w1 * click rate + w2 * conversion rate).

The weight of the click rate w1 and the weight of the conversion rate w2 can be manually set according to the optimized target. For example, the algorithm staff analyzes the data according to the data, and then manually sets the weight of the click rate w1 and the weight of the conversion rate w2 (for example, w1=0.3, w2=0.7) according to the target to be optimized.

In a sorting mechanism optimized for clickthrough rate and conversion rate, each advertiser has a sorting score, which is then sorted by sorting in descending order.

For example, the merchant 1 sort is divided into 0.7, the merchant 2 sort is divided into 0.6, and the merchant 1 rank is in front of the merchant 2. The rankings for Merchant 1 and Merchant 2 are based on a combination of clickthrough rate, conversion rate, and bid.

For another example, the bid of merchant 1 is 1 (here can refer to 1 yuan / click per click), the click rate is 0.8, and the conversion rate is 0.6, then the ranking of merchant 1 is 1* (w1*0.8+w2*0.6 Assume that the bid, click-through rate, and conversion rate of the merchant 2 are 1, 0.6, and 0.6, respectively, and the ranking of the merchant 2 is divided into 1*(0.6*w1+0.6*w2). In the sorting mechanism optimized by click rate and conversion rate, w1 and w2 are the same for all merchants. This way, advertisers can’t optimize individualized appeals with different clickthrough rates and conversion rates.

Platforms that can track transactions will additionally optimize conversion rates. In the three parties (users, advertisers, platforms) involved in the advertising marketing activities, the advertiser's other goals (such as transaction amount, ROI (return on investment)) are not considered in the ranking optimization. The transaction amount in this application refers to the amount of a single order transaction (hereinafter referred to as the transaction amount). Merchants (that is, advertisers) generally want the higher the transaction amount and ROI, the better.

The sorting mechanism optimized by clickthrough rate and conversion rate is not necessarily optimal for advertisers. For example, for an advertiser, if the gross profit of 10 orders converted by 10 users is not as high as the profit of 1 order purchased by another user, the advertiser will be more willing to choose to promote the user. Purchase order. Therefore, conversion rate optimization does not fully meet the advertiser's optimization goals.

In addition, when advertisers are at different stages of development, the appeals for different goals may be different. The sorting mechanism optimized according to click-through rate and the sorting mechanism optimized according to click rate and conversion rate do not consider the different development of advertisers. The appeal of the stage.

As shown in FIG. 1, the sorting method may include the following steps.

In step S110, a plurality of objects to be sorted are acquired according to the request information.

The internet website can present a plurality of objects to the user to enable the user to browse and perform the corresponding conversion operation. For example, in an e-commerce website, an object includes a product recommended to a user, and a user who logs in to the e-commerce website can perform a conversion operation such as further purchase by browsing related information of the product. For example, an application (application, APP) installed on a mobile terminal (for example, a mobile phone, a tablet, a wearable smart device, etc.) can recommend a corresponding store to the user based on the current geographic location of the user or a search keyword input by the user. Shops, such as food and beverage outlets.

Each website or APP can display a plurality of objects based on a certain sorting rule. For example, after the user searches with a search engine, the search results are displayed in a preset sorting manner.

The sorting method provided by the present application is directed to multiple objects, so multiple objects to be sorted are obtained before sorting. For example, when the server of the platform receives the request information, it acquires all objects under the same category as multiple objects to be sorted.

In the present application, an object refers to various products, applications, stores, services, advertisements, and the like that can be presented to a user via the Internet and can be executed by a user, such as a user, an application, a store, a service, an advertisement, and the like.

In an exemplary embodiment, the request information includes search information input by a current user; and/or combined information between the current user and each object; and/or user attributes of the current user.

In an embodiment, the search information may be a search keyword input by the current user, such as "hot pot", "cake", "flower", "mobile business hall around Wangjing" or the like. It should be noted that the search keyword herein is not necessarily the current input by the user, but may also be a search keyword input in the history of the user, or comprehensively consider the keyword currently input by the user and the keyword searched in history. Or search keywords for a period of time in history, such as keywords that were entered last time or last week.

The combination information may be information such as a distance, an orientation, a traffic situation, and the like between the current user and the object, and may also include a matching degree or a correlation between the search keyword input by the current user and the object, such as whether the current user likes the current user. The category of the object, whether it likes the consumption of the business district where the object is located.

In an exemplary embodiment, the user attribute includes any one or more of the current user's taste preferences, environmental preferences, price sensitivity, brand preferences, and the like.

In an embodiment, the user attribute may include personalization information of the current user, such as a taste preference (can be statistically analyzed according to information such as the current user's historical purchase record, order record, etc., for example, the user prefers spicy Sichuan cuisine), environmental preferences (for example, some users value the environment of shopping or eating, hope to spend in quiet stores), price sensitivity (for example, some users may not value the eating environment, pay more attention to cost performance, and another Some users may be price-insensitive), brand preferences (for example, under the same conditions, users are more focused on clothing of a certain brand under the clothing category), category preferences (for example, the user prefers Lu cuisine), and business circle preferences (such as The user is currently in the middle of the two business districts, but the user prefers one of the business districts), distance sensitivity (for example, some users care about the convenience of transportation, and some users only need to eat the food they like. , no matter how far away it is, etc.

Obtaining a plurality of application scenarios for the plurality of objects to be sorted according to the request information. For example, the current user may open a certain APP on the mobile phone, and the homepage of the APP may be based on the current user's user attribute (eg, the user's taste preference) and the combined information between the user and the corresponding object (eg, the The distance between the user and the surrounding stores) acquires the plurality of objects to be sorted. For another example, the current user can open a certain APP on the mobile phone, and the first page of the APP can obtain the plurality of objects to be sorted according to the last week or the last operation record of the current user, for example, an order record. For another example, the store matching the keyword may be acquired as the plurality of objects to be sorted only according to the search keyword input by the current user. For example, the search information of the current user, the combination information between the current user and each object, and the user attribute of the current user may be comprehensively obtained to obtain the plurality of objects to be sorted. This disclosure does not limit this.

In step S120, the sorting feature value of each object is acquired.

The ranking feature value of each of the objects is obtained based on a click feature value, a click weight, a conversion feature value, a conversion weight, a transaction feature value, and a transaction weight of each of the objects.

In an exemplary embodiment, acquiring the sorting feature value of the object includes: acquiring a historical exposure amount, a historical click amount, and a historical order amount of the object; and according to the request information and the historical exposure amount of the object The historical hit quantity obtains a click feature value of the object; and the conversion feature of the object is obtained according to the request information, the click feature value of the object, the historical click amount, and the historical order quantity And acquiring a transaction feature value of the object according to the request information and the conversion feature value of the object; and the click feature value, the conversion feature value, and the transaction feature value according to the object The sorted feature value of the object is calculated.

In an exemplary embodiment, obtaining the click feature value of the object according to the request information and the historical exposure amount and the historical click amount of the object includes: according to the historical exposure amount and location of the object The historical click volume obtains a historical exposure click rate of the object; and obtains a click feature value of the object according to the object attribute of the object and the historical exposure click rate and the request information.

In an embodiment, the first machine model may be used to obtain a click feature value for each object. The first machine model is a model that is trained by the corresponding machine learning algorithm or the deep learning algorithm in advance, and any machine learning algorithm or deep learning algorithm may be used, which is not limited in the disclosure.

In an embodiment, the object attribute may include information such as a category of the object, a business circle, a geographical location, and the like.

It should be noted that the specific descriptions in the foregoing request information, the search information, the combination information, the user attribute, and the object attribute are used for illustration. In different application scenarios, adjustment and setting may be performed according to specific situations.

In an exemplary embodiment, obtaining, according to the request information, the click feature value of the object, the historical click amount, and the historical order quantity, the converted feature value of the object includes: according to the object The historical click amount and the historical order quantity acquire the historical click order rate of the object; and obtain the location according to the object attribute of the object, the historical click order rate and the click feature value, and the request information The transformed feature value of the object.

In this embodiment, the second machine model can be used to obtain the transformed feature value of each object. The second machine model is a model that is trained by using a corresponding machine learning algorithm or a deep learning algorithm in advance, and any machine learning algorithm or deep learning algorithm may be used, which is not limited in the disclosure.

In an exemplary embodiment, obtaining the transaction feature value of the object according to the request information and the converted feature value of the object comprises: obtaining a predicted transaction of the object according to the object attribute of the object and the request information And normalizing the predicted transaction amount of the object by using a preset reference transaction amount; and obtaining the predicted conversion transaction amount of the object and the conversion characteristic value of the object according to the object The transaction feature value of the object.

In this embodiment, the third machine model can be used to obtain the predicted transaction amount of each object. The third machine model is a model that is trained by the corresponding machine learning algorithm or the deep learning algorithm in advance, and any machine learning algorithm or deep learning algorithm may be used, which is not limited in the disclosure.

It should be noted that the first machine model, the second machine model, and the third machine model may respectively adopt different machine learning algorithms or deep learning algorithms, or may use the same machine learning algorithm or deep learning algorithm. Not limited.

In an exemplary embodiment, sorting the plurality of objects according to the sorting feature value of each of the objects comprises: determining, based on the click feature values of the plurality of objects, the clicks a reference click feature value that is normalized by the feature value; a reference conversion feature value for normalizing the transformed feature value is determined based on the transformed feature value of the plurality of objects; The transaction feature value of the object determines a reference click feature value for normalizing the transaction feature value; for each of the objects, the click feature of the object using the reference click feature value Normalizing the value, normalizing the transformed feature value of the object by using the reference conversion feature value, and normalizing the transaction feature value of the object by using the reference transaction feature value, Calculating the ranked feature value of the object based on the normalized click feature value, the normalized transformed feature value, and the normalized transaction feature value of the object; and According to the ranking value for each feature of the object, of the plurality of objects to be sorted.

In an exemplary embodiment, the sorting method further includes: for each of the objects, obtaining a click weight, a conversion weight, and a transaction weight of the object according to a current state of the object.

In an exemplary embodiment, acquiring the click weight, the conversion weight, and the transaction weight of the object according to the current state of the object includes: when the object is in the first state, setting the click right to be greater than the object The conversion weight and the conversion right are greater than the transaction weight; when the object is in the second state, the conversion right is set to be greater than the transaction weight and the transaction right is greater than the click weight for the object; And when the object is in the third state, the transaction right is set to be greater than the conversion weight and the conversion weight is greater than the click weight for the object. The sum of the click weight, the conversion weight, and the transaction weight of the object is a preset constant.

For example, the preset constant may be 1, or may be 2, or any other constant, which is not limited by the disclosure. In the following embodiments, the preset constant is taken as an example.

In an exemplary embodiment, when the object is in the first state, setting the click right to the conversion weight and the conversion weight is greater than the transaction weight for the object includes: when the object is consumed When the budget ratio is within the first preset range, the click weight of the object is increased according to the consumption budget ratio of the object. The consumption budget ratio refers to the ratio of the money paid by the object to the platform and the target budget.

In an exemplary embodiment, when the object is in the second state, setting the conversion right to be equal to the transaction weight and the transaction right is greater than the click weight for the object: when the object is When the consumption budget ratio is within the second preset range, the conversion weight of the object is increased according to the historical exposure conversion rate of the object.

In an exemplary embodiment, when the object is in a third state, setting, for the object, the transaction right is greater than the conversion weight and the conversion weight is greater than the click weight comprises: when the object consumes a budget ratio When in the third preset range, the transaction weight of the object is increased according to the transaction amount of the object.

For example, when the consumption budget ratio of an object is relatively low, the object is considered to be in the first state, and the advertiser is far from obtaining the click rate expected by the budget. When the consumption budget ratio of an object is relatively moderate, the object can be considered to be in the second state, and the conversion weight is optimized. When the consumption budget ratio of an object is relatively high, the object can be considered to be in the third state, and the ROI is low. At this time, the transaction weight can be optimized and the transaction amount of the object can be improved.

In step S130, the plurality of objects are sorted according to the sorting feature value of each object.

In this embodiment, the method may further include: obtaining a ranking score of each object according to the sorting feature value of each object and a bid of the corresponding object; and descending each object according to the sorting score from large to small Arranged, but the disclosure is not limited thereto.

In an exemplary embodiment, the method may further include: outputting the sorted plurality of objects to the client, so that the sorted plurality of objects are displayed on the client.

The ranking method provided by the embodiment obtains the ranking feature value based on the click feature value, the click weight, the conversion feature value, the conversion weight, the transaction feature value and the transaction weight of each object, thereby realizing the advertisement from the perspective of the advertiser optimization target. The main appeal to the optimization of the transaction amount to the sorting mechanism can make the sorting result more reasonable and accurate, and improve the accuracy of the advertising promotion.

As shown in FIG. 2, the sorting method may include the following steps.

In step S210, a plurality of objects to be sorted are acquired according to the request information.

In step S220, the historical exposure amount, the historical click amount, and the historical order quantity of each object are acquired.

In step S230, a click feature value, a conversion feature value, and a transaction feature value of each object are obtained according to the request information, the historical exposure amount, the historical click amount, and the historical order quantity.

In step S240, for each object, the click weight, the conversion weight, and the transaction weight of the object are obtained according to the current state of the object.

In step S250, for each object, the sorting feature value of the object is obtained according to the click feature value, the click weight, the conversion feature value, the conversion weight, the transaction feature value, and the transaction weight of the object.

Here, the plurality of objects are described as a plurality of advertisements placed in a certain platform. When sorting these multiple ads, each ad needs to be assigned a sort score and sorted with this sort score. You can include the goal that the advertiser of the ad wants to optimize when designing the sort score.

For example, the advertisement i, i in the plurality of advertisements represents an advertisement index, and the ranking score is calculated by the following formula (1):

rankScore _i =bid _i ×T _i (1)

In the above formula (1), rankScore _i represents the ranking score of the advertisement i, and bid _i represents the bid of the advertisement i. For example, if the advertisement budget of the advertisement i is 150 yuan per day, and the bid is 1 yuan for one click, the maximum is 150 per day. Second click. T _i represents the ranking feature value of the advertisement i.

Consumption refers to the money that advertisers pay to the platform, and the daily (or weekly or monthly) consumption is less than or equal to the advertising budget for the day (or week or month).

The estimated transaction amount (price _i ) for a user is normalized by the average price of the historical transaction amount in the merchant corresponding to the advertisement i or a historical transaction amount (represented by priceNormDivisor), in this case,

Indicates the relative level of consumption of the user in the merchant corresponding to the advertisement i. α _i represents the click weight of the advertisement i. β _i represents the conversion weight of the advertisement i. γ _i represents the transaction weight of the advertisement i. CTR _i , CVR _i , and price _i respectively represent the estimated exposure click rate, click order rate, and transaction amount of the advertisement i, and the estimated exposure click rate is also referred to as the click feature value, and the ranking feature value T _{i of the} advertisement _i It can be calculated by the following formula (2):

In the above formula (2), priceNormDivisor indicates the average price of a historical transaction amount or a certain historical transaction amount corresponding to the advertisement i. In order to simplify the calculation, for a plurality of advertisements to be sorted, the priceNormDivisor corresponding to each advertisement may be consistent. The CTR _i , CVR _i , and price _i are probability values predicted by the machine learning model according to the historical exposure amount, historical click amount, historical order quantity, historical conversion amount, and historical transaction amount of the advertisement i, respectively. Where CTR _i is the click feature value of the advertisement i, and CTR _i × CVR _i is the conversion feature value of the advertisement i.

The transaction characteristic value for the advertisement i.

Suppose the historical exposure of the advertisement i is 100 times. The exposure here means that the advertisement i is displayed on a website or an APP page, and there are 10 historical clicks, and the historical exposure click rate of the advertisement i is 10/100= 10%; among the 10 historical hits, there is 1 historical order quantity, then the historical click order rate of the advertisement i is 1/10=10%; then the historical exposure conversion rate of the advertisement i is the historical exposure of the advertisement i The click rate is multiplied by the historical click order rate of the advertisement i, that is, 10%×10%=1%. The CTR _i and CVR _i in the above formula are based on historical data (such as historical exposure, historical clicks, historical orders, historical exposure click rate, historical click order rate, historical exposure conversion rate, etc.). ), the store attribute, the user attribute of the current user, the combined information of the store and the current user, etc., the estimated exposure rate and the click order rate of the i-th advertisement. Price _i is a machine model that estimates the transaction amount of the current user, which may be based on the historical transaction amount of the current user, the user attribute of the current user, the store attribute of the store, the combination information of the store and the current user, Data such as the search keyword input by the current user and the matching degree of the store are estimated.

The formula (2) for calculating the ranking feature value T _i of the advertisement i is for a single store corresponding to a single request of the current user. In an embodiment, when the keyword input by the user changes, CTR _i , CVR _i , and price _i change, for example, when the user inputs “hot pot” and “hot pot Wangjing”, the search result ranking is different. . In another embodiment, when the position of the user sends a search request changes, CTR _i , CVR _i , price _i also change, for example, the closer the distance between a store and the user, CTR _i , CVR _i , The price _i is larger; the farther a store is from the user, the smaller the CTR _i , CVR _i , and price _i . Even if the user inputs the same keyword at the same time and in the same place, CTR _i , CVR _i , and price _i are different for different users, because the user attributes of different users are different. Each time a search request is received, the platform's server can estimate by store model or deep learning method based on store attributes, user attributes, and combination information of the store and the user, such as distance, bearing, matching degree, etc. CTR _i , CVR _i , price _i .

For example, for a store corresponding to an advertisement, different users' historical evaluation, rating, star rating, and commentary on the store can be used as the store attribute of the store, and for a specific user, the user's historical evaluation in different stores. , rating, ordering, browsing, collection, etc. can be used to calculate the user's taste preferences, such as some users like hot pot, some users like Sichuan cuisine; user sensitivity to price, such as some users like cheap, environmental preferences For example, some users like user attributes such as good stores.

In formula (2), normalizing the estimated transaction amount can limit the normalized transaction amount to a certain range and facilitate subsequent data processing, so that when priceNormDivisor is consistent with multiple ads to be sorted In the calculation process of the sorting scores, the standards of the stores corresponding to the plurality of advertisements to be sorted are unified.

When determining price _i , a machine learning method well known to those skilled in the art can be employed. For example, normalization by a linear function, the expression (3) of the linear function is as follows:

y=(x-MinValue)/(MaxValue-MinValue) (3)

In the above formula, x and y are the values before and after normalization, and MaxValue and MinValue are the maximum and minimum values of the sample, respectively. For example, when determining price _i , the current user history single transaction amount is between 0-200 yuan, then MaxValue=200, MinValue=0, x represents the current predicted transaction amount, then y represents the normalized transaction. Amount, in this case, x=price _i ,

Then, intelligent analysis calculations are performed for the advertiser's optimization goals. For example, there are different click weights, conversion weights, and transaction weights at different stages of development.

The first stage of development is for newly opened stores (also known as new stores). The new store urgently needs to obtain exposure, click volume, and conversion volume. At this time, it is possible to focus on optimizing the click rate, that is, the corresponding advertisement of this part of the store. α _i can be higher, β _{i is} second, and γ _{i is the} smallest. For example, α _i may be 0.5, β _i may be 0.4, and γ _i is 0.1.

In an embodiment, judging whether a store is a new store, a new store or an old store may be judged according to the length of opening of the store. For example, two time thresholds, that is, a first time threshold and a second time threshold, may be preset, the first time threshold being less than the second time threshold. The store whose opening time is less than the first time threshold is a new store. The store whose opening time is between the first time threshold and the second time threshold is between the old and new stores. The store with an opening time greater than the second time threshold is an old store. It should be noted that the first time threshold and the second time threshold may be a floating value, and one month and two weeks are all ok, and may be set according to a specific application scenario.

The second stage of development is for stores that have gained popularity (also known as developing mid-stores). Because of the popularity of developing stores, advertisers can use idle resources (idle resources refer to the difference between advertising budget and consumption) to optimize the transaction amount. For the development of stores, focus on optimizing the conversion rate and transaction volume of advertising i, therefore, with respect to new store β _{_i,} γ _i, developing stores β _{_i,} γ _i can be higher.

In an exemplary embodiment, the ROI may have different definitions in different applications, such as ROI = sales profit/advertising cost * 100%, or ROI = advertising drain profit / advertising cost. Among them, the profit of advertising drainage = total profit - natural single profit, natural single profit refers to the profit brought by the platform to the merchant without advertising; the total investment of advertising = online input + offline input + activity input. For example, to calculate the ROI of an e-commerce platform on a certain day, it is ROI=(total cash profit of the order generated on the day-average total profit per day on the inactive day)/(online input + offline input + other input).

It should be noted that, in this embodiment, according to the definition of the ROI in different occasions, the denominator (for example, the advertising cost) in the formula for calculating the ROI is also unchanged when the advertiser budget is unchanged. In addition, since the profit and the transaction amount are positively correlated, the profit is optimized while optimizing the transaction amount, so the ROI value is optimized while optimizing the transaction amount.

In one embodiment, for a developing store, the two parameters β _μ and γ _i can be adjusted by a “consumption/advertising budget” (also referred to as a consumption budget ratio). For example, when 70% of the budget is used, the β _i and γ _{i of the} store can be set to 0.45 and 0.45 respectively, and when the budget ratio is 80%, the β _i and γ _{i of the} store can be set to 0.4 and 0.5, respectively.

The third stage of development is for an old store with an opening time greater than the second time threshold. Assuming that the idle resources of the old store have been exhausted, a reflection indicator in the advertising system is the day/week/month/year (this is also a parameter of the time range) The advertising budget that can be adjusted according to the specific application scenario has been exhausted. For this part of the business, you can increase the transaction amount and increase the ROI while keeping its investment (consumption) unchanged. In this case, the γ _i weight can be set to be the highest, and β _{i is} second.

For the above analysis, it is possible to analyze the historical data of each store, determine the current stage of each store, and set the click weight, conversion weight and transaction weight according to the corresponding stage of the store.

Ideally, it is possible to divide each store according to the length of the store, whether it is a new store, a new store or an old store, but in reality, the store on the platform may be operated for a period of time before advertising. If the store is running well, even if it is just put into advertising, the consumption budget will be higher. Therefore, it is also possible to use the size of the budget to judge whether it is a new store, a new store or an old store.

The following describes how to calculate α _i , β _i and γ _i by taking the consumption budget ratio as an example. α _i +β _i +γ _i =1.

In the above formula (2), suppose CTR _i = 0.3, CVR _i = 0.3,

In this case, the click eigenvalue is 0.3, the conversion eigenvalue is 0.09, the transaction eigenvalue is 0.0027, the product of the conversion weight and the conversion eigenvalue, and the product of the transaction weight and the transaction eigenvalue may be much smaller than the click weight and click feature. The product of the values. To further optimize the experience of the merchant (that is, the advertiser), you can analyze the CTR, CTR × CVR of multiple ads to be sorted,

Approximate data distribution, normalized by functions that match this distribution.

Taking the CTR of multiple advertisements to be sorted as an example, assuming that the CTR of multiple advertisements to be sorted is between 0.05 and 0.5, the whole data is normally distributed, and the median is 0.2, and CTR _i = (0.3) can be calculated in the following manner. -0.2) / (0.5 - 0.05) = 0.22.

Similarly, CTR × CVR,

The data conforms to a distribution function that can be normalized by the median of the distribution function.

In another embodiment, the mean value of the click feature values of the plurality of ads to be sorted (also referred to as reference click feature values), the mean value of the converted feature values (also referred to as reference conversion feature values), and the transaction characteristics may also be analyzed. The mean of the values (also known as the reference transaction eigenvalue). In this case, the click feature value of the advertisement i is normalized by using the reference click feature value to obtain a normalized click feature value, and the conversion feature value of the advertisement i is normalized by using the reference conversion feature value. A normalized transformed feature value is obtained, and the transaction feature value of the advertisement i is normalized by using the reference transaction feature value to obtain a normalized transaction feature value. Thus, the normalized click feature value of the advertisement i, the normalized conversion feature value, and the normalized transaction feature value are on the order of magnitude. The ranking score of the advertisement i is based on the normalized click feature value, the normalized conversion feature value, and the normalized transaction feature value.

It should be noted that α _i , β _i and γ _i can be based on the previous online experiment.

with

Make adjustments. However, the present disclosure is not limited to this. The previous online experiment refers to the process of calculating the ranking score of the advertisement i.

For the different optimization goals of advertisers, the previous online experiment

with

The adjustment factor is a _i , b _i , c _i .

Optimize exposure and click to increase α _i for merchants or stores with a consumption budget ratio lower than 50% (this value is a value that can be set independently according to the specific situation). Adjustment factor of α _i

The consumption budget ratio can be determined by using, for example, when the advertisement i corresponds to the median consumption budget of the store corresponding to the store corresponding to the plurality of advertisements to be sorted, the adjustment coefficient of α _i is calculated as follows

For example, for the store corresponding to the advertisement i, the previous online experiment

with

Both are 1/3,

with

The adjustment coefficients a _i , b _i and c _i are both 1, and the following formula (4) can be obtained:

When the median of the budget consumption ratio of the stores corresponding to the plurality of advertisements to be sorted is 80% and the consumption budget ratio of the store is 40%, the adjustment coefficient of α _i

For 2, the following formula (5) is obtained:

At this time, the value on the right side of the equation (5) is not equal to 1. In this case, the equal sign on both sides of the formula (5) can be divided by 4/3 to obtain the formula (6):

under these circumstances,

with

Thus α _{i is} greater than

In this case, if

then

and

It should be noted that the adjustment coefficient of α _i is calculated in the above embodiment.

It is based on the consumption budget of the store corresponding to the multiple advertisements to be sorted divided by the median consumption budget ratio of the store, but the quantile of the median can be adjusted, for example, adjusted to 30% quantile.

The median (Median, also known as the median) represents a value in the middle of a sample, population, or probability distribution. For a limited number set, one of the positive middles can be found as the median by sorting all the observations. If there are even numbers of observations, the average of the two most intermediate values can be taken as the median.

The quantile is a value that is in the same position after arranging all the data in order of size. For example, there are 1000 values that have been sorted by size, and the 30% quantile is the 300th value. There are 1000 values that have been sorted by size, and the 20% quantile is the 200th value.

For merchants or stores that have a high budget (for example, between 50% and 90%, which is an adjustable range), optimize the conversion rate to increase β _i .

The historical exposure conversion rate here is equal to the historical exposure click rate multiplied by the historical click order rate. In this case, if

then

and

For merchants with a particularly high budget (eg >90%, which is an adjustable parameter), optimize the ROI to increase γ _i . In an embodiment,

In another embodiment,

In this case, if

then

and

In step S260, the plurality of objects are sorted according to the sorting feature value of each object.

The sorting method provided by the present disclosure starts with the optimization goal of the merchant, not only optimizes the user experience, but also optimizes the merchant experience.

In addition, in the sorting method provided by the present disclosure, different advertisers have different development demands on the amount of clicks, conversion amount, transaction amount, etc. at different stages of development, and can be set for each of the three different target targets for each advertiser. Different weights are optimized. In addition to considering three factors of click volume, conversion rate and transaction amount, the method of the embodiment of the present disclosure can intelligently analyze the characteristics of different merchants and optimize the target, and set the weight of the merchant differentiation for each different optimization target of each merchant. Each advertiser, according to its own development stage, several optimization goals (click volume, conversion volume, transaction amount) can each have different weights, and can be unified into the same sorting mechanism, which can optimize the merchant experience. In some embodiments, when the advertiser's bid is unchanged, when the user inputs the request information on the platform, the user's possible transaction amount is estimated according to the relevant data, and the order of the advertisement is optimized based on the user's possible transaction amount. The score, when the sort score increases, the number of times the advertiser gets the transaction amount may increase in a certain period of time, in which case the advertiser's ROI is improved. The sorting method of the present application does not require too much advertiser participation (such as actively setting the target to be optimized), and it can intelligently analyze and determine the target to be optimized by the advertiser.

It should be noted that the sorting method shown in FIG. 1 and FIG. 2 can be applied to the server. The server can be a server or a cloud server.

It should be noted that the sorting apparatus shown in FIG. 3 can be applied to the server. The server can be a server or a cloud server. The disclosure is not limited in this disclosure.

As shown in FIG. 3, the sorting apparatus 100 provided in this embodiment may include: an object obtaining module 110, a feature value obtaining module 120, and a sorting module 130.

The object obtaining module 110 may be configured to acquire a plurality of objects to be sorted according to the request information.

The feature value acquisition module 120 can be configured to acquire a sort feature value of each object.

The sorting module 130 can be configured to sort the plurality of objects according to the sorted feature values of each of the objects.

The sorting feature value of each object is obtained based on the click feature value, the click weight, the conversion feature value, the conversion weight, the transaction feature value, and the transaction weight of each object.

In an exemplary embodiment, the request information includes search information input by a current user and/or combined information between the current user and each object and/or user attributes of the current user.

In an exemplary embodiment, the user attribute includes any one or more of the current user's taste preference, environmental preference, price sensitivity, and brand preference.

In an exemplary embodiment, the feature value acquisition module 120 may include a historical data acquisition sub-module and a feature value calculation sub-module. The historical data acquisition sub-module may be configured to acquire a historical exposure amount, a historical click amount, and a historical order quantity of the object. The feature value calculation sub-module may be configured to obtain a click feature value of the object according to the request information and the historical exposure amount of the object and the historical click amount; according to the request information and the object Obtaining a conversion feature value of the object by the click feature value, the historical click amount, and the historical order quantity; acquiring the transaction feature of the object according to the request information and the conversion feature value of the object And calculating the sorting feature value of the object according to the click feature value, the transformed feature value, and the transaction feature value of the object.

In an exemplary embodiment, the feature value calculation sub-module may include an exposure click-through rate calculation unit and a click feature value calculation unit. The exposure click rate calculation unit may be configured to acquire a historical exposure click rate of the object according to the historical exposure amount of the object and the historical click amount. The click feature value calculation unit may be configured to obtain a click feature value of the object according to the object attribute of the object and the historical exposure click rate and the request information.

In an exemplary embodiment, the feature value calculation sub-module may include a click order rate calculation unit and a conversion feature value calculation unit. The click order rate calculation unit may be configured to acquire a historical click order rate of the object according to the historical click amount of the object and the historical order quantity. The conversion feature value calculation unit may be configured to obtain a conversion feature value of the object according to the object attribute of the object, the historical click order rate, the click feature value, and the request information.

In an exemplary embodiment, the feature value calculation sub-module may include a transaction amount estimation unit, a normalization unit, and a transaction feature value calculation unit. The transaction amount estimating unit may be configured to obtain a predicted transaction amount of the object according to the object attribute of the object and the request information. The normalization unit may be configured to normalize the predicted transaction amount of the object by using a preset reference transaction amount. The transaction feature value calculation unit may be configured to obtain a transaction feature value of the object according to the normalized predicted transaction amount of the object and the converted feature value of the object.

In an exemplary embodiment, the sorting device may further include a weight acquisition module. The weight obtaining module may be configured to obtain, for each of the objects, a click weight, a conversion weight, and a transaction weight of the object according to the current state of the object.

In an exemplary embodiment, the weight acquisition module may include a first weight setting sub-module, a second weight setting sub-module, and a third weight setting sub-module. The first weight setting sub-module may be configured to set the click right of the object to be greater than the conversion weight and the conversion weight is greater than the transaction weight when the object is in the first state. The second weight setting sub-module may be configured to set the conversion right of the object to be greater than the transaction weight and the transaction right is greater than the click weight when the object is in the second state. The third weight setting sub-module may be configured to set the transaction right of the object to be greater than the conversion weight and the conversion weight is greater than the click weight when the object is in the third state. The sum of the click weight, the conversion weight, and the transaction weight of each object is a preset constant.

In an exemplary embodiment, the first weight setting sub-module may include a first weight calculation unit. The first weight calculation unit may be configured to increase a click weight of the object according to a consumption budget ratio of the object when the consumption budget ratio of the object is within the first preset range.

In an exemplary embodiment, the second weight setting sub-module may include a second weight calculation unit. The second weight calculation unit may be configured to increase the conversion weight of the object according to the historical exposure conversion rate of the object when the consumption budget ratio of the object is within the second preset range.

In an exemplary embodiment, the third weight setting sub-module may include a third weight calculation unit. The third weight calculation unit may be configured to increase the transaction weight of the object according to the transaction amount of the object when the consumption budget ratio of the object is within the third preset range.

It should be noted that the specific implementation of the module of the sorting device may refer to the content of the sorting method shown in FIG. 1 and FIG. 2 above, and details are not described herein again.

According to an exemplary embodiment of the present disclosure, there is also provided an electronic device, which may include a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the program is processed by the processor The method steps shown in FIG. 1 or FIG. 2 above are implemented when the device is executed.

Referring now to Figure 4, there is shown a block diagram of an electronic device 400 suitable for use in implementing the embodiments of the present application. The electronic device shown in FIG. 4 is merely an example, and should not impose any limitation on the function and scope of use of the embodiments of the present application.

As shown in FIG. 4, electronic device 400 includes a processor 401 that can perform various appropriate actions and processes in accordance with programs stored in memory 403. In particular, the processes described above with reference to the flowcharts may be implemented as a computer software program in accordance with an embodiment of the present disclosure. For example, an embodiment of the present disclosure includes a computer program product comprising a computer program carried on a computer readable medium, the computer program comprising program code for executing the method illustrated in the flowchart, wherein the computer program is processed When the 401 is executed, the above-described functions defined in the system of the present application are executed. The processor 401, the memory 403, and the communication interface 402 are connected to each other through a bus.

The flowchart and block diagrams in the Figures illustrate the architecture, functionality, and operation of possible implementations of terminals, servers, methods and computer program products in accordance with various embodiments of the present application. In this regard, each block of the flowchart or block diagrams can represent a module, a program segment, or a portion of code that includes one or more Executable instructions. It should also be noted that in some alternative implementations, the functions noted in the blocks may also occur in a different order than that illustrated in the drawings. For example, two successively represented blocks may in fact be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending upon the functionality involved. It is also noted that each block of the block diagrams or flowcharts, and combinations of blocks in the block diagrams or flowcharts, can be implemented by a dedicated hardware-based system that performs the specified function or operation, or can be used A combination of dedicated hardware and computer instructions is implemented.

The modules involved in the embodiments of the present application may be implemented by software or by hardware. The described modules can also be arranged in the processor.

In another aspect, the present disclosure also provides a computer readable medium, which may be included in the apparatus described in the above embodiments, or may be separately present and not incorporated into the apparatus. The computer readable medium carries one or more programs, and when the one or more programs are executed by the device, the device includes: acquiring a plurality of objects to be sorted according to the request information; acquiring a sorting feature of each object a value; and sorting the plurality of objects according to the sorted feature value of each object; wherein the sorting feature value of each object is based on a click feature value, a click weight, a conversion feature value, a conversion weight, The transaction feature value and the transaction weight are obtained.

The exemplary embodiments of the present disclosure have been specifically shown and described above. It should be understood that the invention is not limited to the details of the details of the embodiments of the invention.

Claims

A sorting method that includes:

Obtaining a plurality of objects to be sorted according to the request information;

Obtaining a sorting feature value for each of the objects;

Sorting the plurality of objects according to the sorting feature value of each of the objects;

The ranking feature value of each of the objects is obtained based on a click feature value, a click weight, a conversion feature value, a conversion weight, a transaction feature value, and a transaction weight of each of the objects.
The method of claim 1, obtaining the ordered feature values of the object comprises:

Obtaining historical exposure, historical hits, and historical orders for the object;

Obtaining a click feature value of the object according to the request information and the historical exposure amount of the object and the historical click amount;

Obtaining a conversion feature value of the object according to the request information, the click feature value of the object, the historical click amount, and the historical order quantity;

Obtaining a transaction feature value of the object according to the request information and the conversion feature value of the object; and

And calculating the sorting feature value of the object according to the click feature value, the conversion feature value, and the transaction feature value of the object.
The method according to claim 2, wherein obtaining the click feature value of the object according to the request information and the historical exposure amount and the historical click amount of the object comprises:

Obtaining a historical exposure click rate of the object according to the historical exposure amount of the object and the historical click amount; and

Obtaining a click feature value of the object according to the object attribute of the object and the historical exposure click rate and the request information.
The method according to claim 2, wherein obtaining the transformed feature value of the object according to the request information, the click feature value of the object, the historical click amount, and the historical order quantity comprises:

Obtaining a historical click order rate of the object according to the historical click amount of the object and the historical order quantity; and

And converting the feature value of the object according to the object attribute of the object, the historical click order rate, the click feature value, and the request information.
The method according to claim 2, wherein obtaining the transaction feature value of the object according to the request information and the converted feature value of the object comprises:

Obtaining a predicted transaction amount of the object according to the object attribute of the object and the request information;

Normalizing the predicted transaction amount of the object using a preset reference transaction amount; and

Obtaining a transaction feature value of the object according to the normalized predicted transaction amount of the object and the transformed feature value of the object.
The method of claim 1, sorting the plurality of objects according to the sorted feature value of each of the objects, comprising:

Determining a reference click feature value for normalizing the click feature value based on the click feature value of the plurality of objects;

Determining a reference conversion feature value for normalizing the transformed feature value based on the transformed feature values of the plurality of objects;

Determining a reference click feature value for normalizing the transaction feature value based on the transaction feature value of the plurality of objects;

For each of the objects,

Normalizing the click feature value of the object by using the reference click feature value,

Normalizing the transformed feature values of the object using the reference transformed feature values,

Normalizing the transaction feature values of the object using the reference transaction feature values,

Calculating the ranked feature value of the object based on the normalized click feature value, the normalized transformed feature value, and the normalized transaction feature value of the object; and

The plurality of objects are sorted according to the sorted feature value of each of the objects.
The method of claim 1 further comprising:

For each of the objects, the click weight, conversion weight, and transaction weight of the object are obtained according to the current state of the object.
The method according to claim 7, wherein obtaining the click weight, the conversion weight, and the transaction weight of the object according to the current state of the object includes:

When the object is in the first state, the click right is set to be greater than the conversion weight and the conversion right is greater than the transaction weight for the object;

And when the object is in the second state, setting the conversion right to be equal to the transaction weight and the transaction right is greater than the click weight for the object; and

And when the object is in the third state, the transaction right is set to be greater than the conversion weight and the conversion right is greater than the click weight for the object;

The sum of the click weight, the conversion weight, and the transaction weight of the object is a preset constant.
The method according to claim 8, when the object is in the first state, setting the click right to the conversion weight and the conversion weight is greater than the transaction weight for the object:

When the consumption budget ratio of the object is within the first preset range, the click weight of the object is increased according to the consumption budget ratio of the object.
The method according to claim 8, wherein when the object is in the second state, setting the conversion right to be equal to the transaction weight and the transaction right is greater than the click weight for the object comprises:

When the consumption budget ratio of the object is within the second preset range, the conversion weight of the object is increased according to the historical exposure conversion rate of the object.
The method according to claim 8, when the object is in the third state, setting the transaction right to be greater than the conversion weight for the object, and the conversion weight is greater than the click weight comprises:

When the consumption budget ratio of the object is within the third preset range, the transaction weight of the object is increased according to the transaction amount of the object.
The method according to any one of claims 1 to 11, wherein the request information comprises one or more of the following:

Search information entered by the current user;

Combined information between the current user and each object; and

The user attribute of the current user.
A sorting device comprising:

An object obtaining module, configured to acquire, according to the request information, multiple objects to be sorted;

An eigenvalue obtaining module, configured to obtain a sorting feature value of each of the objects;

a sorting module, configured to sort the plurality of objects according to the sorting feature value of each of the objects;

The ranking feature value of each of the objects is obtained based on a click feature value, a click weight, a conversion feature value, a conversion weight, a transaction feature value, and a transaction weight of each of the objects.
An electronic device comprising a memory, a processor, and a computer program stored on the memory and operable on the processor, wherein the program is executed by the processor to implement any of claims 1-12 The method described.
A computer readable medium having stored thereon a computer program, wherein the program is executed by a processor to implement the method of any of claims 1-12.