CN110599295B - Method, device and equipment for pushing articles - Google Patents
Method, device and equipment for pushing articles Download PDFInfo
- Publication number
- CN110599295B CN110599295B CN201910780017.6A CN201910780017A CN110599295B CN 110599295 B CN110599295 B CN 110599295B CN 201910780017 A CN201910780017 A CN 201910780017A CN 110599295 B CN110599295 B CN 110599295B
- Authority
- CN
- China
- Prior art keywords
- conversion rate
- article
- estimated
- users
- user
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0251—Targeted advertisements
- G06Q30/0269—Targeted advertisements based on user profile or attribute
- G06Q30/0271—Personalized advertisement
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/50—Network services
- H04L67/55—Push-based network services
Abstract
The embodiment of the specification provides a method, a device and equipment for pushing an article. Inputting the user characteristics and the attribute characteristics of the first article into a conversion rate estimation model to output the estimated conversion rate of the current user for the first article. And determining the real conversion rate of the first item by the plurality of users according to the proportion of conversion users who perform the preset operation on the first item in the process of exposing the first item to the plurality of users. And obtaining the estimated conversion rate of each user in the plurality of users to the first article through the conversion rate estimation model, thereby determining the average value of the estimated conversion rate of the plurality of users to the first article. And adjusting the estimated conversion rate of the current user for the first article based on the real conversion rate and the average value of the estimated conversion rate. And determining whether to push the first article to the current user or not based on the adjusted estimated conversion rate.
Description
Technical Field
One or more embodiments of the present disclosure relate to the field of computer technologies, and in particular, to a method, an apparatus, and a device for pushing an article.
Background
In recent years, recommendation technology has been rapidly developed in the fields of e-commerce, advertising and the like, and immeasurable value is created for the recommendation technology.
Existing recommendation techniques are typically implemented based on a recommendation model. The recommendation model is trained based on historical operating data of a large number of recommended articles. As in the e-commerce field, recommendation models may be pre-trained based on purchase records of large quantities of merchandise. Then, in the process that the user visits the E-commerce website, the goods preferred by the user can be identified based on the recommendation model, and the goods can be recommended to the user.
However, for a new item, since the amount or proportion of the historical operating data corresponding to the new item is below a threshold, there is often a large bias when implementing its recommendation based on a recommendation model. Therefore, it is necessary to provide a more accurate method for pushing new articles.
Disclosure of Invention
One or more embodiments of the present specification describe a method, an apparatus, and a device for pushing an article, which can implement accurate pushing of a new article.
In a first aspect, a method for pushing an article is provided, including:
acquiring user characteristics of a current user;
inputting the user characteristics and the attribute characteristics of the first article into a conversion rate pre-estimation model to output the pre-estimated conversion rate of the current user for the first article; the estimated conversion rate represents the possibility that a current user performs a preset operation on the first article; the conversion rate pre-estimation model is obtained by training based on historical operation data of articles, wherein the quantity or the proportion of the historical operation data corresponding to the first article is lower than a threshold value;
determining the real conversion rate of the first item by the plurality of users according to the proportion of conversion users who perform the predetermined operation on the first item in the process of exposing the first item to the plurality of users;
obtaining the estimated conversion rate of each user in the plurality of users to the first article through the conversion rate estimation model, thereby determining the average value of the estimated conversion rate of the plurality of users to the first article;
based on the real conversion rate and the average value of the estimated conversion rate, adjusting the estimated conversion rate of the current user for the first article;
and determining whether to push the first article to the current user or not based on the adjusted estimated conversion rate.
In a second aspect, there is provided an article pushing device, comprising:
the device comprises an acquisition unit, a processing unit and a processing unit, wherein the acquisition unit is used for acquiring the user characteristics of a current user;
the input unit is used for inputting the user characteristics acquired by the acquisition unit and the attribute characteristics of the first article into a conversion rate pre-estimation model so as to output the pre-estimated conversion rate of the current user for the first article; the estimated conversion rate represents the possibility that a current user performs a preset operation on the first article; the conversion rate pre-estimation model is obtained by training based on historical operation data of articles, wherein the quantity or the proportion of the historical operation data corresponding to the first article is lower than a threshold value;
the determining unit is used for determining the real conversion rate of the first item for the plurality of users according to the proportion of conversion users who perform the predetermined operation on the first item in the process of exposing the first item to the plurality of users;
the determining unit is further configured to obtain, through the conversion rate pre-estimation model, pre-estimation conversion rates of the users for the first item, so as to determine a mean value of the pre-estimation conversion rates of the users for the first item;
the adjusting unit is used for adjusting the estimated conversion rate of the current user for the first article based on the real conversion rate determined by the determining unit and the average value of the estimated conversion rates;
the determining unit is further configured to determine whether to push the first item to the current user based on the estimated conversion rate adjusted by the adjusting unit.
In a third aspect, there is provided an article pushing apparatus, including:
a memory;
one or more processors; and
one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the programs when executed by the processors implement the steps of:
acquiring user characteristics of a current user;
inputting the user characteristics and the attribute characteristics of the first article into a conversion rate pre-estimation model to output the pre-estimated conversion rate of the current user for the first article; the estimated conversion rate represents the possibility that a current user performs a preset operation on the first article; the conversion rate pre-estimation model is obtained by training based on historical operation data of articles, wherein the quantity or the proportion of the historical operation data corresponding to the first article is lower than a threshold value;
determining the real conversion rate of the first item by the plurality of users according to the proportion of conversion users who perform the predetermined operation on the first item in the process of exposing the first item to the plurality of users;
obtaining the estimated conversion rate of each user in the plurality of users to the first article through the conversion rate estimation model, thereby determining the average value of the estimated conversion rate of the plurality of users to the first article;
based on the real conversion rate and the average value of the estimated conversion rate, adjusting the estimated conversion rate of the current user for the first article;
and determining whether to push the first article to the current user or not based on the adjusted estimated conversion rate.
According to the pushing method, the pushing device and the pushing equipment for the article, which are provided by one or more embodiments of the specification, the user characteristics of the current user are obtained. Inputting the user characteristics and the attribute characteristics of the first article into a conversion rate estimation model to output the estimated conversion rate of the current user for the first article. The estimated conversion rate characterizes a likelihood that a current user will perform a predetermined operation on the first item. The conversion rate estimation model is obtained by training based on historical operation data of the articles, wherein the quantity or the proportion of the historical operation data corresponding to the first article is lower than a threshold value. And determining the real conversion rate of the first item by the plurality of users according to the proportion of conversion users who perform the preset operation on the first item in the process of exposing the first item to the plurality of users. And obtaining the estimated conversion rate of each user in the plurality of users to the first article through the conversion rate estimation model, thereby determining the average value of the estimated conversion rate of the plurality of users to the first article. And adjusting the estimated conversion rate of the current user for the first article based on the real conversion rate and the average value of the estimated conversion rate. And determining whether to push the first article to the current user or not based on the adjusted estimated conversion rate. That is, in the solution provided in this specification, for the first item, the conversion rate of the user for the first item is estimated based on the conversion rate estimation model. And then, based on the real conversion rate of the plurality of users to the first article and the average value of the estimated conversion rates, adjusting the estimated conversion rate of the current user to the first article so as to enable the estimated conversion rate of the current user to the first article to be close to the real value. Therefore, the accuracy of the estimated conversion rate can be greatly improved, and the accurate pushing of new articles can be further realized.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present disclosure, and it is obvious for those skilled in the art to obtain other drawings based on the drawings without creative efforts.
Fig. 1 is a schematic view of an application scenario of a pushing method for an article provided in the present specification;
FIG. 2 is a flow chart of a method for training a conversion prediction model provided herein;
fig. 3 is a flowchart of a pushing method for an item according to an embodiment of the present disclosure;
FIG. 4 is a schematic view of a pushing device for pushing an item provided in one embodiment of the present disclosure;
fig. 5 is a schematic diagram of an article pushing apparatus according to an embodiment of the present disclosure.
Detailed Description
The scheme provided by the specification is described below with reference to the accompanying drawings.
Before describing the solution provided in the present specification, the inventive concept of the present solution will be explained below.
Firstly, the estimated conversion rate of the current user for the first article is output based on the conversion rate estimation model. The conversion rate estimation model is obtained by training based on historical operation data of the articles. The history operation may include, but is not limited to, a browsing operation, a clicking operation, a purchasing operation, and the like. It is noted that the amount or proportion of the historical operational data corresponding to the first item is below a threshold. In this specification, the first item may also be referred to as a new item. In addition, the estimated conversion rate is indicative of a likelihood that a current user will perform a predetermined operation on the first item. The predetermined operation herein may include, but is not limited to, a click operation or a purchase operation.
Specifically, the user characteristics of the current user and the attribute characteristics of the first item may be input into the conversion rate prediction model to output the predicted conversion rate of the current user for the first item. And then, determining the real conversion rate and the average value of the estimated conversion rate of the first article by a plurality of users, and adjusting the estimated conversion rate based on the real conversion rate and the average value. Adjustments herein include, but are not limited to, translational adjustments. For example, the difference between the real conversion rate and the average of the estimated conversion rate may be determined, and the difference may be used to measure the difference between the estimated conversion rate and the real value. And based on the difference value of the two, carrying out translation adjustment on the estimated conversion rate so that the estimated conversion rate of the user for the first article can be close to a real value. Finally, it is determined whether to push the first item to the user based on the adjusted estimated conversion rate.
The above is the inventive concept provided in the present specification, and the scheme provided in the present specification can be obtained based on the inventive concept, and the detailed description of the scheme is provided below.
Fig. 1 is a schematic view of an application scenario of a pushing method for an article provided in this specification. In fig. 1, the stream data processing platform 102 is configured to count the exposure number and the conversion number of the first item in real time based on the exposure log and the conversion log of the first item. The exposure log can be used to record user information exposed to the user and information such as exposure time. The conversion log can be used for recording information such as conversion users and conversion time. The recommendation platform 104 is configured to output, based on a conversion rate pre-estimation model trained in advance, respective pre-estimated conversion rates corresponding to a plurality of predetermined articles by a current user, so as to obtain a plurality of pre-estimated conversion rates. For a first item of the plurality of items, the recommendation platform 104 may first determine the real conversion rate and the average of the estimated conversion rates of the plurality of users for the first item, and then adjust the estimated conversion rate of the current user for the first item based on the real conversion rate and the average of the estimated conversion rates. Finally, the recommendation platform 104 may sort the plurality of articles based on the estimated conversion rates corresponding to the plurality of articles or the adjusted estimated conversion rates, and push the top N articles in the front of the sorting to the current user.
Based on the above-described inventive concept, the present solution first outputs the estimated conversion rate of the current user for the first item based on the conversion rate estimation model. Therefore, before describing the present solution, the training process of the conversion rate estimation model is described as follows:
FIG. 2 is a flowchart of a method for training a conversion rate estimation model provided in the present specification. As shown in fig. 2, the method may include:
at step 202, a batch of samples is collected.
At step 204, a positive sample is determined from the batch of samples.
The positive sample includes a user characteristic of the first user and a first tag value. In addition, the positive sample may also include attribute characteristics of the sample item. The first tag value characterizes a predetermined operation performed on the sample item by the first user.
Specifically, after exposing the sample item to a plurality of users, if any user performs a predetermined operation on the sample item, the any user may be the first user. The predetermined operation may include, but is not limited to, a click operation or a purchase operation. For example, the predetermined operation may be a purchase operation. For example, the sample article is an advertisement, and the predetermined operation may be a click operation.
The user characteristics may be obtained based on operation behavior data of the first user recorded in advance, or may be preset, and specific contents thereof will be described later. In one example, with respect to the first tag value, it may be the first user's true conversion rate (also referred to as the true value) for the sample item. Since the first user performed a predetermined operation on the sample article, the true conversion rate thereof is 100%, and thus the first tag value may be 1.
And step 206, taking the samples except the positive sample in the batch of samples as negative samples.
The negative examples include a user characteristic of the second user and a second tag value, the second tag value characterizing that the second user has not performed a predetermined operation on the example item.
Specifically, after the sample item is exposed to a plurality of users, any one of the users who does not perform the predetermined operation on the sample item may be used as the second user. That is, although the second user does not perform the predetermined operation on the sample item, the second user performs the browsing operation on the sample item.
In the above-mentioned example, since the second user does not perform the predetermined operation on the sample article, the true conversion rate thereof is 0%, and thus the above-mentioned second tag value may be 0.
At step 208, negative samples are sampled such that the number of negative samples sampled satisfies a predetermined ratio with the number of positive samples.
It should be noted that, in practice, the number of collected negative samples is much larger than that of positive samples, and therefore, for the collected negative samples, it is common to sample them, i.e. to discard part of the negative samples, so as to ensure the reasonability of the positive and negative samples.
Step 210, training a conversion rate estimation model based on the positive sample and the sampled negative sample.
The transformation ratio estimation model may be obtained by training a Logistic Regression (LR) model and/or a Deep Neural Network (DNN) model or a Gradient Boosting Decision Tree (GBDT) model based on a positive sample and a sampled negative sample.
The above is a description of a training method of a conversion rate estimation model, and after the model is obtained by training, the pushing method of the article provided in this specification may be executed.
Fig. 3 is a flowchart of a pushing method for an item according to an embodiment of the present disclosure. The execution subject of the method may be a device with processing capabilities: a server or system or device, such as recommendation system 104 in fig. 1. As shown in fig. 3, the method may specifically include:
step 302, obtaining the user characteristics of the current user.
User characteristics herein may include, but are not limited to, user attribute characteristics, user profile characteristics, and the like. The user attribute characteristics may include, for example, the age, gender, occupation, and education of the user, among others. The user image features may be features obtained based on a feature model, such as purchasing power and purchasing preference of the user, and the like.
Step 304, inputting the user characteristics and the attribute characteristics of the first article into the conversion rate estimation model to output the estimated conversion rate of the current user for the first article.
The first item herein may include, but is not limited to, a new good or a new advertisement, etc. For the new product as an example, the attribute features may include, but are not limited to, product identification, name, brand, category, and the like.
It should be noted that, as can be seen from the foregoing description of the conversion rate prediction model, the solution provided in the present specification does not train the model based on all collected negative samples. Therefore, the difference between the estimated conversion rate of the current user for the first article and the true value based on the model output is large. Therefore, the above estimated conversion can be calibrated (Calibration) so that its absolute magnitude can approach the true value.
Specifically, the estimated conversion rate may be calibrated by using an order-preserving regression algorithm (Isotonic regression) or a negative sampling reduction method.
Taking the negative sampling reduction method as an example, the calibration process can be realized by the following formula: pCVR _ calibrated ═ pCVR/(pCVR + (1-pCVR)/rate _ neg), where pCVR _ calibrated is the estimated conversion rate after calibration, pCVR is the estimated conversion rate before calibration, and rate _ neg is the negative sampling ratio. The negative sample ratio here may be determined based on the number of negative samples sampled and the total number of negative samples.
It is understood that the principle of calibrating the estimated conversion rate by the negative sampling reduction method can be as follows: and enabling the currently obtained estimated conversion rate to be close to the conversion rate output by the model obtained by training all the collected negative samples.
And step 306, determining the real conversion rate of the first item by the plurality of users according to the proportion of conversion users who perform the predetermined operation on the first item in the plurality of users during the process of exposing the first item to the plurality of users.
It will be appreciated that this true conversion is also a target for subsequent adjustment of the estimated conversion.
For the above true conversion, the determination can be as follows: the number of exposures and the number of conversions for the first item are obtained. The number of exposures may be the total number of the plurality of users, and the number of conversions may be the total number of the conversion users. It is judged whether the exposure number is less than a predetermined exposure amount. If so, the number of exposures and the number of conversions are corrected based on a predetermined exposure amount. Based on the corrected exposure number and the corrected conversion number, a true conversion rate of the first article is determined. If not, determining the true conversion rate of the first article based on the exposure number and the conversion number.
In this specification, the setting of the above-described predetermined exposure amount may satisfy the following preconditions: the true conversion rate can be reflected. Therefore, when the exposure number is less than the predetermined exposure amount, it may be considered that it is not sufficient to reflect the true conversion rate, and it may be corrected. Further, when the exposure amount is less than the predetermined exposure amount, the inversion amount thereof may be considered to also not satisfy the condition, and therefore the inversion amount may also be corrected based on the predetermined exposure amount.
In one implementation, the correction process of the exposure number and the conversion number may be as follows: the conversion number correction amount is determined based on the predetermined exposure amount and the overall average conversion rate. The overall average conversion here is determined based on the number of exposures of a predetermined number of articles and the number of conversions. Specifically, the respective exposure numbers and the conversion numbers may be counted for each of a predetermined number of articles. And summing all the counted exposure numbers, summing all the counted conversion numbers, and putting the summed conversion numbers in the summed exposure numbers to obtain the overall average conversion rate. The predetermined exposure amount is multiplied by the entire conversion to obtain a conversion number correction amount. Then, based on the preset exposure, correcting the exposure number to obtain a corrected exposure number; the conversion number is corrected based on the conversion number correction amount to obtain a corrected conversion number.
In other implementations, the above-described conversion number correction amount may not be determined or may be set to an empirical value. When the amount of correction of the inversion number is not determined, the exposure number and the inversion number may be corrected based on a predetermined exposure amount, and the present specification does not limit the amount.
In addition, for the above-mentioned correction of the exposure number based on the predetermined exposure amount and the correction of the conversion number based on the conversion number correction amount, a specific implementation manner may be: the exposure count is added to a predetermined exposure amount to obtain a corrected exposure count. The converted number is added to the converted number correction amount to obtain a corrected converted number.
In view of the above, one way to determine true conversion may be:
real_CVR＝(#convert+M)/(#expo+N),when#expo<N
real_CVR＝#convert/#expo,when#expo>＝N
where real _ CVR is the true conversion, # expo is the number of exposures, # convert is the number of conversions, N is the predetermined amount of exposure, and M is the conversion correction.
Of course, the actual conversion rate can also be determined by other determination methods according to the above description, which is not listed in the present specification.
It is understood that, in the embodiment of the present specification, when the exposure number is determined to be smaller than the predetermined exposure amount, after the exposure number and the conversion number are corrected based on the predetermined exposure amount or the predetermined exposure amount and the conversion number correction amount, the accuracy of the acquired true conversion rate can be greatly improved.
It should be noted that, the above is a description of the true conversion rate of the first item by the multiple users, and the following is a description of the determination process of the average value of the estimated conversion rates of the first item by the multiple users.
And 308, obtaining the estimated conversion rate of each user in the plurality of users to the first article through the conversion rate estimation model, thereby determining the average value of the estimated conversion rates of the plurality of users to the first article.
Specifically, for each user in the plurality of users, the user characteristics of the user and the attribute characteristics of the first item are input into the conversion rate prediction model, so that the predicted conversion rate of each user in the plurality of users for the first item is output. And averaging or weighted averaging the estimated conversion rate of each user of the plurality of users to the first article to determine the average value of the estimated conversion rate of the plurality of users to the first article.
It can be understood that, similarly to the above-mentioned estimated conversion rate of the current user for the first item, there is also a difference between the estimated conversion rate of each of the multiple users for the first item and the real value obtained here, and therefore, in an implementation manner, before performing the above-mentioned averaging or weighted averaging, the estimated conversion rate of each of the multiple users for the first item may be calibrated, and a specific calibration method thereof is as described above, and is not repeated here.
When the estimated conversion rates of the users for the first item are calibrated, the average value of the estimated conversion rates of the users for the first item may be obtained by averaging or weighting the average values of the calibrated estimated conversion rates.
In another implementation, the estimated conversion rate of each of the plurality of users for the first item may be averaged or weighted-averaged, and then the result of the averaging or weighted-averaging may be calibrated, which is not limited in this specification.
And 310, adjusting the estimated conversion rate of the current user for the first article based on the real conversion rate and the average value of the estimated conversion rate.
In one implementation, the translation may be determined based on a difference between the true conversion and the mean of the predicted conversions. The translation amount represents the difference degree between the estimated conversion rate and the true value of the current user for the first article. And performing translation adjustment on the estimated conversion rate of the current user on the first article based on the determined translation amount.
In one example, the above translation adjustment process can be expressed as the following equation:
pCVR_Adjustment＝pCVR_calibrated+(real_CVR-avg_pCVR_calibrated)
the pCVR-Adjustment is the estimated conversion rate after the translation Adjustment, the pCVR-calibrated is the estimated conversion rate of the current user for the first article, the real-CVR is the real conversion rate of the plurality of users for the first article, and the avg-pCVR-calibration is the average value of the estimated conversion rates of the plurality of users for the first article. real _ CVR-avg _ pCVR _ constrained is the amount of translation.
It is understood that when the calibration operation for the current user's estimated conversion rate for the first item is not performed, the pCVR _ calibrated may be replaced with the current user's pre-calibrated estimated conversion rate for the first item.
It can be seen from the above that, in the implementation manner, the estimated conversion rate output by the model is subjected to translation adjustment to obtain the final conversion rate of the user for the first article, and different estimated conversion rates are output by the model for different users, that is, the estimated conversion rate output by the model can reflect the personalized information of the corresponding user, so that the conversion rate after translation adjustment can also reflect the personalized information. Therefore, when pushing the articles, different articles can be pushed for different users based on the conversion rate after translation adjustment, and personalized delivery of the articles can be realized.
In another implementation, the estimated conversion rate of the first item by the current user may also be adjusted based on a ratio between the actual conversion rate and the average value of the estimated conversion rate, which is not limited in this specification.
Step 312, determining whether to push the first item to the current user based on the adjusted estimated conversion rate.
For example, when there are a plurality of predetermined articles, the plurality of predetermined articles may be sorted in an order from a large estimated conversion rate to a small estimated conversion rate corresponding to each of the plurality of predetermined articles by the current user. And then pushing the top N items ranked in the front to the user. Wherein N is a positive integer.
In summary, the method for pushing an article provided in the embodiment of the present specification may determine a difference between an estimated conversion rate and a real value of a current user for a first article based on a difference between a real conversion rate of a plurality of users for the first article and an average time of the estimated conversion rate. And based on the difference degree, carrying out translation adjustment on the estimated conversion rate of the current user on the first article. Therefore, the personalized delivery of the new articles can be realized, and the flow of the new articles can be regulated and controlled to a reasonable range. In addition, the scheme does not need field professional knowledge as a precondition, does not require good generalization of the attributes of the articles, and has wider application range. In addition, the user does not need to perform early feedback collection, and the overall service effect is better guaranteed.
Corresponding to the above method for pushing an article, an embodiment of the present specification further provides an article pushing device, as shown in fig. 4, the device may include:
an obtaining unit 402, configured to obtain a user characteristic of a current user.
An input unit 404, configured to input the user characteristic acquired by the acquisition unit 402 and the attribute characteristic of the first item into the conversion rate prediction model, so as to output a predicted conversion rate of the current user for the first item. The estimated conversion rate characterizes a likelihood that a current user will perform a predetermined operation on the first item. The conversion rate estimation model is obtained by training based on historical operation data of the articles, wherein the quantity or the proportion of the historical operation data corresponding to the first article is lower than a threshold value.
A determining unit 406, configured to determine a true conversion rate of the first item by the multiple users according to a proportion of conversion users performing the predetermined operation on the first item in the multiple users during the process of exposing the first item to the multiple users.
The determining unit 406 may specifically be configured to:
the number of exposures and the number of conversions for the first item are obtained. Here, the exposure number is the total number of the plurality of users, and the conversion number is the total number of the conversion users.
It is judged whether the exposure number is less than a predetermined exposure amount.
If so, the number of exposures and the number of conversions are corrected based on a predetermined exposure amount. Based on the corrected exposure number and the corrected conversion number, a true conversion rate of the first article is determined.
If not, determining the true conversion rate of the first article based on the exposure number and the conversion number.
The determining unit 406 may further specifically be configured to:
the conversion number correction amount is determined based on the predetermined exposure amount and the overall average conversion rate. The overall average conversion is determined based on the number of exposures of the predetermined number of articles and the number of conversions.
The exposure count is corrected based on a predetermined exposure amount to obtain a corrected exposure count.
The conversion number is corrected based on the conversion number correction amount to obtain a corrected conversion number.
The determining unit 406 is further configured to obtain, through the conversion rate prediction model, the predicted conversion rate of each user of the multiple users for the first item, so as to determine a mean value of the predicted conversion rates of the multiple users for the first item.
The determining unit 406 may further specifically be configured to:
for each user in the plurality of users, inputting the user characteristics of the user and the attribute characteristics of the first article into the conversion rate prediction model so as to output the predicted conversion rate of each user in the plurality of users for the first article.
And averaging or weighted averaging the estimated conversion rate of each user in the plurality of users to the first article to determine the average value of the estimated conversion rate of the plurality of users to the first article.
The determining unit 406 may further specifically be configured to:
and calibrating the estimated conversion rate of each user in the plurality of users on the first article by adopting a sequence-preserving regression algorithm or based on the sampling proportion of the negative sample to obtain a plurality of calibrated estimated conversion rates. The sampling proportion is determined based on the number of negative samples sampled and the total number of negative samples.
Averaging or weighted averaging the plurality of calibrated estimated conversion rates to determine an average value of the plurality of users estimated conversion rates for the first item.
An adjusting unit 408, configured to adjust the estimated conversion rate of the first item for the current user based on the real conversion rate determined by the determining unit 406 and the average value of the estimated conversion rates.
The adjusting unit 408 may specifically be configured to:
and determining the translation amount based on the difference value between the real conversion rate and the estimated conversion rate mean value. The translation amount represents the difference degree between the estimated conversion rate and the true value of the current user for the first article.
And performing translation adjustment on the estimated conversion rate of the current user on the first article based on the determined translation amount.
The determining unit 406 is further configured to determine whether to push the first item to the current user based on the estimated conversion rate adjusted by the adjusting unit 408.
Optionally, the apparatus may further include:
a collection unit (not shown) for collecting a batch of samples.
The determining unit 406 is further configured to determine a positive sample from the batch of samples collected by the collecting unit. The positive sample includes a user characteristic of the first user and a first tag value, the first tag value characterizing that the first user performed a predetermined operation on the sample item.
And the determining unit 406 is further configured to take samples other than the positive sample in the batch of samples collected by the collecting unit as negative samples. The negative examples include a user characteristic of the second user and a second label value. The second tag value characterizes that the second user has not performed the predetermined operation on the sample item.
A sampling unit (not shown in the figure) for sampling the negative samples determined by the determination unit 406 such that the number of sampled negative samples and the number of sampled positive samples satisfy a predetermined ratio.
And a training unit (not shown in the figure) for training the conversion rate estimation model based on the positive samples determined by the determination unit 406 and the negative samples sampled by the acquisition unit.
Optionally, the apparatus may further include:
and the calibration unit (not shown in the figure) is used for calibrating the estimated conversion rate of the first article of the current user by adopting a sequence-preserving regression algorithm or based on the sampling proportion of the negative samples. The sampling ratio here is determined based on the number of negative samples sampled and the total number of negative samples.
The adjusting unit 408 may specifically be configured to:
and adjusting the estimated conversion rate after calibration based on the real conversion rate and the average value of the estimated conversion rate.
The functions of each functional module of the device in the above embodiments of the present description may be implemented through each step of the above method embodiments, and therefore, a specific working process of the device provided in one embodiment of the present description is not repeated herein.
In the pushing device for an item provided in an embodiment of the present specification, the obtaining unit 402 obtains a user characteristic of a current user. The input unit 404 inputs the user characteristic and the attribute characteristic of the first item into the conversion rate estimation model to output the estimated conversion rate of the current user for the first item. The estimated conversion rate characterizes a likelihood that a current user will perform a predetermined operation on the first item. The conversion rate estimation model is obtained by training based on historical operation data of the articles, wherein the quantity or the proportion of the historical operation data corresponding to the first article is lower than a threshold value. The determining unit 406 determines the real conversion rate of the first item by the plurality of users according to the proportion of conversion users performing the predetermined operation on the first item in the plurality of users during the process of exposing the first item to the plurality of users. The determining unit 406 obtains the estimated conversion rate of each of the plurality of users for the first item through the conversion rate estimation model, so as to determine the estimated conversion rate of the plurality of users for the first item. The average adjusting unit 408 adjusts the estimated conversion rate of the first item for the current user based on the real conversion rate and the estimated conversion rate average. The determining unit 406 determines whether to push the first item to the current user based on the adjusted estimated conversion rate. From this, can realize the accurate propelling movement of new article.
The pushing device for the article provided by one embodiment of the present specification may be a module or unit of the recommendation platform 104 in fig. 1.
Correspondingly to the pushing method of the article, an embodiment of the present specification further provides an article pushing apparatus, as shown in fig. 5, the apparatus may include: memory 502, one or more processors 504, and one or more programs. Wherein the one or more programs are stored in the memory 502 and configured to be executed by the one or more processors 504, the programs when executed by the processors 504 implement the steps of:
and acquiring the user characteristics of the current user.
Inputting the user characteristics and the attribute characteristics of the first article into a conversion rate estimation model to output the estimated conversion rate of the current user for the first article. The estimated conversion rate characterizes a likelihood that a current user will perform a predetermined operation on the first item. The conversion rate estimation model is obtained by training based on historical operation data of the articles, wherein the quantity or the proportion of the historical operation data corresponding to the first article is lower than a threshold value.
And determining the real conversion rate of the first item by the plurality of users according to the proportion of conversion users who perform the preset operation on the first item in the process of exposing the first item to the plurality of users.
And obtaining the estimated conversion rate of each user in the plurality of users to the first article through a conversion rate estimation model, thereby determining the average value of the estimated conversion rate of the plurality of users to the first article. And adjusting the estimated conversion rate of the current user for the first article based on the real conversion rate and the average value of the estimated conversion rate.
And determining whether to push the first article to the current user or not based on the adjusted estimated conversion rate.
The pushing device for the articles, provided by one embodiment of the specification, can realize accurate pushing of new articles.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the apparatus embodiment, since it is substantially similar to the method embodiment, the description is relatively simple, and for the relevant points, reference may be made to the partial description of the method embodiment.
The steps of a method or algorithm described in connection with the disclosure herein may be embodied in hardware or may be embodied in software instructions executed by a processor. The software instructions may consist of corresponding software modules that may be stored in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor such the processor can read information from, and write information to, the storage medium. Of course, the storage medium may also be integral to the processor. The processor and the storage medium may reside in an ASIC. Additionally, the ASIC may reside in a server. Of course, the processor and the storage medium may reside as discrete components in a server.
Those skilled in the art will recognize that, in one or more of the examples described above, the functions described in this invention may be implemented in hardware, software, firmware, or any combination thereof. When implemented in software, the functions may be stored on or transmitted over as one or more instructions or code on a computer-readable medium. Computer-readable media includes both computer storage media and communication media including any medium that facilitates transfer of a computer program from one place to another. A storage media may be any available media that can be accessed by a general purpose or special purpose computer.
The foregoing description has been directed to specific embodiments of this disclosure. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.
The above-mentioned embodiments, objects, technical solutions and advantages of the present specification are further described in detail, it should be understood that the above-mentioned embodiments are only specific embodiments of the present specification, and are not intended to limit the scope of the present specification, and any modifications, equivalent substitutions, improvements and the like made on the basis of the technical solutions of the present specification should be included in the scope of the present specification.
Claims (17)
1. A pushing method of an article comprises the following steps:
acquiring user characteristics of a current user;
inputting the user characteristics and the attribute characteristics of the first article into a conversion rate pre-estimation model to output the pre-estimated conversion rate of the current user for the first article; the estimated conversion rate represents the possibility that a current user performs a preset operation on the first article; the conversion rate estimation model is obtained by training based on historical operation data of articles, wherein the quantity of the historical operation data corresponding to the first article is lower than a threshold value;
determining the real conversion rate of the first item by the plurality of users according to the proportion of conversion users who perform the predetermined operation on the first item in the process of exposing the first item to the plurality of users;
obtaining the estimated conversion rate of each user in the plurality of users to the first article through the conversion rate estimation model, thereby determining the average value of the estimated conversion rate of the plurality of users to the first article;
based on the real conversion rate and the average value of the estimated conversion rate, adjusting the estimated conversion rate of the current user for the first article;
and determining whether to push the first article to the current user or not based on the adjusted estimated conversion rate.
2. The method of claim 1, the determining the true conversion rates of the plurality of users for the first item, comprising:
acquiring the exposure number and the conversion number of the first article; the exposure number is a total number of the plurality of users, and the conversion number is a total number of the conversion users;
judging whether the exposure number is less than a preset exposure amount;
if yes, correcting the exposure number and the conversion number based on the preset exposure amount; determining a true conversion rate of the first item based on the corrected exposure number and the corrected conversion number;
if not, determining the real conversion rate of the first article based on the exposure number and the conversion number.
3. The method of claim 2, wherein the correcting the exposure count and the conversion count based on the predetermined exposure includes:
determining a conversion number correction amount based on the predetermined exposure amount and the overall average conversion rate; the overall average conversion rate is determined based on the number of exposures of the predetermined number of articles and the number of conversions;
correcting the exposure number based on the preset exposure amount to obtain the corrected exposure number;
and correcting the conversion number based on the conversion number correction amount to obtain the corrected conversion number.
4. The method of claim 1, wherein the conversion rate pre-estimation model is trained by the following steps:
collecting a batch of samples;
determining a positive sample from the batch of samples; the positive sample includes a user characteristic of a first user and a first tag value; the first tag value characterizes that the first user performed a predetermined operation on a sample item;
taking samples in the batch of samples except the positive samples as negative samples; the negative examples include a user characteristic of a second user and a second tag value; the second tag value characterizes that the second user has not performed a predetermined operation on a sample item;
sampling the negative samples so that the number of the sampled negative samples and the number of the sampled positive samples meet a preset proportion;
and training the conversion rate estimation model based on the positive samples and the sampled negative samples.
5. The method of claim 4, further comprising, prior to said adjusting the current user's estimated conversion for the first item based on the real conversion and the average of the estimated conversions:
calibrating the estimated conversion rate of the current user on the first article by adopting a sequence-preserving regression algorithm or based on the sampling proportion of the negative sample to obtain the calibrated estimated conversion rate; the sampling proportion is determined based on the number of negative samples sampled and the total number of negative samples;
the adjusting the estimated conversion rate of the current user for the first article based on the real conversion rate and the estimated conversion rate average value comprises:
and adjusting the estimated conversion rate after calibration based on the real conversion rate and the average value of the estimated conversion rate.
6. The method of claim 4, wherein said determining a mean of said plurality of users' estimated conversion rates for said first item comprises:
for each user in the plurality of users, inputting the user characteristics of the user and the attribute characteristics of the first article into the conversion rate pre-estimation model so as to output the pre-estimated conversion rate of each user in the plurality of users for the first article;
and averaging or weighted averaging the estimated conversion rates of the users to the first item to determine the average value of the estimated conversion rates of the users to the first item.
7. The method of claim 6, the averaging or weighted averaging of the estimated conversion rate of each of the plurality of users for the first item, comprising:
calibrating the estimated conversion rate of each user in the plurality of users on the first article by adopting a sequence preserving regression algorithm or based on the sampling proportion of the negative sample to obtain a plurality of calibrated estimated conversion rates; the sampling proportion is determined based on the number of negative samples sampled and the total number of negative samples;
averaging or weighted averaging the plurality of calibrated estimated conversion rates to determine an average of the plurality of users' estimated conversion rates for the first item.
8. The method of claim 1, wherein adjusting the estimated conversion rate of the first item for the current user based on the real conversion rate and the average estimated conversion rate comprises:
determining translation amount based on the difference value between the real conversion rate and the estimated conversion rate mean value; the translation amount represents the difference degree between the estimated conversion rate and the true value of the current user for the first article;
and performing translation adjustment on the estimated conversion rate of the current user on the first article based on the determined translation amount.
9. A pushing device of an item, comprising:
the device comprises an acquisition unit, a processing unit and a processing unit, wherein the acquisition unit is used for acquiring the user characteristics of a current user;
the input unit is used for inputting the user characteristics acquired by the acquisition unit and the attribute characteristics of the first article into a conversion rate pre-estimation model so as to output the pre-estimated conversion rate of the current user for the first article; the estimated conversion rate represents the possibility that a current user performs a preset operation on the first article; the conversion rate estimation model is obtained by training based on historical operation data of articles, wherein the quantity of the historical operation data corresponding to the first article is lower than a threshold value;
the determining unit is used for determining the real conversion rate of the first item for the plurality of users according to the proportion of conversion users who perform the predetermined operation on the first item in the process of exposing the first item to the plurality of users;
the determining unit is further configured to obtain, through the conversion rate pre-estimation model, pre-estimation conversion rates of the users for the first item, so as to determine a mean value of the pre-estimation conversion rates of the users for the first item;
the adjusting unit is used for adjusting the estimated conversion rate of the current user for the first article based on the real conversion rate determined by the determining unit and the average value of the estimated conversion rates;
the determining unit is further configured to determine whether to push the first item to the current user based on the estimated conversion rate adjusted by the adjusting unit.
10. The apparatus according to claim 9, wherein the determining unit is specifically configured to:
acquiring the exposure number and the conversion number of the first article; the exposure number is a total number of the plurality of users, and the conversion number is a total number of the conversion users;
judging whether the exposure number is less than a preset exposure amount;
if yes, correcting the exposure number and the conversion number based on the preset exposure amount; determining a true conversion rate of the first item based on the corrected exposure number and the corrected conversion number;
if not, determining the real conversion rate of the first article based on the exposure number and the conversion number.
11. The apparatus of claim 10, the determining unit further specifically configured to:
determining a conversion number correction amount based on the predetermined exposure amount and the overall average conversion rate; the overall average conversion rate is determined based on the number of exposures of the predetermined number of articles and the number of conversions;
correcting the exposure number based on the preset exposure amount to obtain the corrected exposure number;
and correcting the conversion number based on the conversion number correction amount to obtain the corrected conversion number.
12. The apparatus of claim 9, further comprising:
the collecting unit is used for collecting a batch of samples;
the determining unit is further used for determining a positive sample from the batch of samples collected by the collecting unit; the positive sample includes a user characteristic of a first user and a first tag value; the first tag value characterizes that the first user performed a predetermined operation on a sample item;
the determining unit is further configured to take samples other than the positive sample in the batch of samples collected by the collecting unit as negative samples; the negative examples include a user characteristic of a second user and a second tag value; the second tag value characterizes that the second user has not performed a predetermined operation on a sample item;
the sampling unit is used for sampling the negative samples determined by the determining unit, so that the number of the sampled negative samples and the number of the sampled positive samples meet a preset proportion;
and the training unit is used for training the conversion rate estimation model based on the positive sample determined by the determining unit and the negative sample sampled by the sampling unit.
13. The apparatus of claim 12, further comprising:
the calibration unit is used for calibrating the estimated conversion rate of the current user on the first article by adopting a sequence-preserving regression algorithm or based on the sampling proportion of the negative sample so as to obtain the estimated conversion rate after calibration; the sampling proportion is determined based on the number of negative samples sampled and the total number of negative samples;
the adjusting unit is specifically configured to:
and adjusting the estimated conversion rate after calibration based on the real conversion rate and the average value of the estimated conversion rate.
14. The apparatus of claim 12, the determining unit further specifically configured to:
for each user in the plurality of users, inputting the user characteristics of the user and the attribute characteristics of the first article into the conversion rate pre-estimation model so as to output the pre-estimated conversion rate of each user in the plurality of users for the first article;
and averaging or weighted averaging the estimated conversion rates of the users to the first item to determine the average value of the estimated conversion rates of the users to the first item.
15. The apparatus of claim 14, the determining unit further specifically configured to:
calibrating the estimated conversion rate of each user in the plurality of users on the first article by adopting a sequence preserving regression algorithm or based on the sampling proportion of the negative sample to obtain a plurality of calibrated estimated conversion rates; the sampling proportion is determined based on the number of negative samples sampled and the total number of negative samples;
averaging or weighted averaging the plurality of calibrated estimated conversion rates to determine an average of the plurality of users' estimated conversion rates for the first item.
16. The apparatus according to claim 9, wherein the adjusting unit is specifically configured to:
determining translation amount based on the difference value between the real conversion rate and the estimated conversion rate mean value; the translation amount represents the difference degree between the estimated conversion rate and the true value of the current user for the first article;
and performing translation adjustment on the estimated conversion rate of the current user on the first article based on the determined translation amount.
17. A pushing device of an item, comprising:
a memory;
one or more processors; and
one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the programs when executed by the processors implement the steps of:
acquiring user characteristics of a current user;
inputting the user characteristics and the attribute characteristics of the first article into a conversion rate pre-estimation model to output the pre-estimated conversion rate of the current user for the first article; the estimated conversion rate represents the possibility that a current user performs a preset operation on the first article; the conversion rate estimation model is obtained by training based on historical operation data of articles, wherein the quantity of the historical operation data corresponding to the first article is lower than a threshold value;
determining the real conversion rate of the first item by the plurality of users according to the proportion of conversion users who perform the predetermined operation on the first item in the process of exposing the first item to the plurality of users;
obtaining the estimated conversion rate of each user in the plurality of users to the first article through the conversion rate estimation model, thereby determining the average value of the estimated conversion rate of the plurality of users to the first article;
based on the real conversion rate and the average value of the estimated conversion rate, adjusting the estimated conversion rate of the current user for the first article;
and determining whether to push the first article to the current user or not based on the adjusted estimated conversion rate.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910780017.6A CN110599295B (en) | 2019-08-22 | 2019-08-22 | Method, device and equipment for pushing articles |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910780017.6A CN110599295B (en) | 2019-08-22 | 2019-08-22 | Method, device and equipment for pushing articles |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110599295A CN110599295A (en) | 2019-12-20 |
| CN110599295B true CN110599295B (en) | 2022-04-08 |
Family
ID=68855286
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910780017.6A Active CN110599295B (en) | 2019-08-22 | 2019-08-22 | Method, device and equipment for pushing articles |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110599295B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111259222B (en) * | 2020-01-22 | 2023-08-22 | 北京百度网讯科技有限公司 | Article recommendation method, system, electronic equipment and storage medium |
| CN111798280B (en) * | 2020-09-08 | 2020-12-15 | 腾讯科技（深圳）有限公司 | Multimedia information recommendation method, device and equipment and storage medium |
| CN115330428A (en) * | 2021-05-11 | 2022-11-11 | 腾讯科技（深圳）有限公司 | Data calibration method and device, computer equipment and readable storage medium |
| CN117114821A (en) * | 2023-10-23 | 2023-11-24 | 湖南快乐阳光互动娱乐传媒有限公司 | Article recommendation method and device, storage medium and electronic equipment |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105446802A (en) * | 2014-08-13 | 2016-03-30 | 阿里巴巴集团控股有限公司 | Operation execution method and device based on conversion rate |
| CN105760400A (en) * | 2014-12-19 | 2016-07-13 | 阿里巴巴集团控股有限公司 | Method and device for ranking push messages based on search behavior |
| CN109903111A (en) * | 2017-12-11 | 2019-06-18 | 北京京东尚科信息技术有限公司 | For the sort method of personalized recommendation, order models training method and ordering system |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170046732A1 (en) * | 2015-08-14 | 2017-02-16 | International Business Machines Corporation | Training a machine to dynamically determine and communicate customized, product-dependent promotions with no or limited historical data over a network |
| US11115492B2 (en) * | 2018-02-13 | 2021-09-07 | Ebay, Inc. | Methods and system for determining parameters for a product based on analysis of input media |
-
2019
- 2019-08-22 CN CN201910780017.6A patent/CN110599295B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105446802A (en) * | 2014-08-13 | 2016-03-30 | 阿里巴巴集团控股有限公司 | Operation execution method and device based on conversion rate |
| CN105760400A (en) * | 2014-12-19 | 2016-07-13 | 阿里巴巴集团控股有限公司 | Method and device for ranking push messages based on search behavior |
| CN109903111A (en) * | 2017-12-11 | 2019-06-18 | 北京京东尚科信息技术有限公司 | For the sort method of personalized recommendation, order models training method and ordering system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110599295A (en) | 2019-12-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110599295B (en) | Method, device and equipment for pushing articles | |
| CN107423442B (en) | Application recommendation method and system based on user portrait behavior analysis, storage medium and computer equipment | |
| CN110503531B (en) | Dynamic social scene recommendation method based on time sequence perception | |
| CN107016026B (en) | User tag determination method, information push method, user tag determination device, information push device | |
| CN108665311B (en) | Electric commercial user time-varying feature similarity calculation recommendation method based on deep neural network | |
| JP2011248829A (en) | Evaluation prediction device, evaluation prediction method and program | |
| CN109685537B (en) | User behavior analysis method, device, medium and electronic equipment | |
| CN111783810B (en) | Method and device for determining attribute information of user | |
| CN111738534A (en) | Training of multi-task prediction model, and prediction method and device of event type | |
| US20220405806A1 (en) | Utilizing machine-learning models to create target audiences with customized auto-tunable reach and accuracy | |
| CN112380449B (en) | Information recommendation method, model training method and related device | |
| Wang et al. | Modeling uncertainty to improve personalized recommendations via Bayesian deep learning | |
| Stoklosa et al. | Heterogeneous capture–recapture models with covariates: A partial likelihood approach for closed populations | |
| CN109460474B (en) | User preference trend mining method | |
| JP7423694B2 (en) | Machine learning methods, training methods, prediction systems, and non-transitory computer-readable media | |
| JP7441270B2 (en) | Machine learning methods, training methods, prediction systems, and non-transitory computer-readable media | |
| CN115423040A (en) | User portrait identification method and AI system of interactive marketing platform | |
| CN111400512B (en) | Method and device for screening multimedia resources | |
| CN110245990B (en) | Advertisement recommendation method and device, electronic equipment and storage medium | |
| JP7431890B2 (en) | Machine learning methods, training methods, prediction systems, and non-transitory computer-readable media | |
| JP2019109662A (en) | Classification device, data structure, classification method, and program | |
| JP7429260B2 (en) | Machine learning methods, training methods, prediction systems, and non-transitory computer-readable media | |
| CN117035873B (en) | Multi-task combined prediction method for few-sample advertisement | |
| CN116071129A (en) | Commodity recommendation method and device, storage medium and electronic equipment | |
| US20220284300A1 (en) | Techniques to tune scale parameter for activations in binary neural networks |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 40019448 Country of ref document: HK |
|
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20200924 Address after: Cayman Enterprise Centre, 27 Hospital Road, George Town, Grand Cayman Islands Applicant after: Innovative advanced technology Co.,Ltd. Address before: Cayman Enterprise Centre, 27 Hospital Road, George Town, Grand Cayman Islands Applicant before: Advanced innovation technology Co.,Ltd. Effective date of registration: 20200924 Address after: Cayman Enterprise Centre, 27 Hospital Road, George Town, Grand Cayman Islands Applicant after: Advanced innovation technology Co.,Ltd. Address before: A four-storey 847 mailbox in Grand Cayman Capital Building, British Cayman Islands Applicant before: Alibaba Group Holding Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |