CN112818231A

CN112818231A - Information delivery method and device, electronic equipment and storage medium

Info

Publication number: CN112818231A
Application number: CN202110130826.XA
Authority: CN
Inventors: 刘焦
Original assignee: Beijing QIYI Century Science and Technology Co Ltd
Current assignee: Beijing QIYI Century Science and Technology Co Ltd
Priority date: 2021-01-29
Filing date: 2021-01-29
Publication date: 2021-05-18

Abstract

The embodiment of the invention provides an information delivery method, an information delivery device, electronic equipment and a storage medium. The information delivery method comprises the following steps: obtaining interest parameters of a target user on information categories; obtaining a sequencing result of the information to be released corresponding to the target user; adjusting the sorting result of the information to be put according to the interest parameters of the target user on the information category to obtain the adjusted sorting result; and according to the adjusted sorting result, performing information delivery on the target user. The embodiment of the invention combines the interest of the user to carry out information delivery, can improve the relevance between the delivered information and the user, and thus improves the accuracy of information delivery.

Description

Information delivery method and device, electronic equipment and storage medium

Technical Field

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

Background

With the rapid development of internet technology, users increasingly rely on obtaining various information through networks. In order to meet the needs of users, various video websites come along, and the video websites can provide various information such as videos for the users. In order to promote the commodities, each merchant usually releases information to the user, so that the user can purchase recommended commodities more conveniently.

With the impact of the internet and digital transmission technology on the whole society, information delivery becomes more difficult, and accurate delivery of information is the trend of information popularization or marketing at present. Currently, in the process of information delivery, estimation is usually performed based on click rate and conversion rate of a large number of users. However, in practical situations, the information delivered may not be really needed by the user. Therefore, the accuracy of the current information delivery is low, and the delivery effect is poor.

Disclosure of Invention

The embodiment of the invention aims to provide an information delivery method, an information delivery device, electronic equipment and a storage medium, so as to improve the accuracy of information delivery. The specific technical scheme is as follows:

in a first aspect of the present invention, an information delivery method is provided, including:

obtaining interest parameters of a target user on information categories;

obtaining a sequencing result of the information to be released corresponding to the target user;

adjusting the sorting result of the information to be put according to the interest parameters of the target user on the information category to obtain the adjusted sorting result;

and according to the adjusted sorting result, performing information delivery on the target user.

In a second aspect of the present invention, there is also provided an information delivery apparatus, including:

the first acquisition module is used for acquiring interest parameters of a target user on information categories;

the second acquisition module is used for acquiring the sequencing result of the information to be released corresponding to the target user;

the adjusting module is used for adjusting the sorting result of the information to be put according to the interest parameters of the target user on the information category to obtain the adjusted sorting result;

and the releasing module is used for releasing the information of the target user according to the adjusted sequencing result.

In another aspect of the present invention, there is also provided an electronic device, including a processor, a communication interface, a memory and a communication bus, where the processor, the communication interface, and the memory complete communication with each other through the communication bus; a memory for storing a computer program; and the processor is used for realizing any one of the information delivery methods when executing the program stored in the memory.

In yet another aspect of the present invention, there is also provided a computer-readable storage medium, which stores instructions that, when executed on a computer, cause the computer to implement any one of the information delivery methods described above.

In yet another aspect of the present invention, there is also provided a computer program product containing instructions which, when run on a computer, cause the computer to implement any of the information delivery methods described above.

According to the information delivery method, the information delivery device, the electronic equipment and the storage medium provided by the embodiment of the invention, when information delivery is carried out, the target user is subjected to information delivery according to the adjusted sorting result by acquiring the interest parameters of the target user in the information category and adjusting the sorting result of the information to be delivered corresponding to the target user according to the interest parameters of the target user in the information category. As described above, in the prior art, the click rate and the conversion rate of a large number of users are estimated, and the information of the large number of users is uniformly analyzed, so that the delivered information can reflect the preference degree of the large number of users to the message to a certain extent, rather than the information actually required by a single user; in contrast, in the embodiment of the invention, information delivery is carried out by combining the interest parameters of the single users on the information categories, and the sequencing result of the information to be delivered can be adjusted based on the information categories, so that the information delivered by any single user can be more inclined to the related information of the interest categories of the single user to a certain extent, and the relevance between the delivered information and the user can be improved, thereby improving the accuracy of information delivery, delivering the information to people who really have needs, optimizing the experience of the audience, reducing the disturbance to non-target people, and enhancing the information delivery effect.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below.

Fig. 1 is a flowchart illustrating steps of an information delivery method according to an embodiment of the present invention.

Fig. 2 is a flowchart of steps of a prediction model building process in the embodiment of the present invention.

Fig. 3 is a block diagram of an information delivery apparatus according to an embodiment of the present invention.

Fig. 4 is a block diagram of another information delivery apparatus according to an embodiment of the present invention.

Fig. 5 is a block diagram of an electronic device in an embodiment of the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be described below with reference to the drawings in the embodiments of the present invention.

The information delivery can be applied to various scenes such as a process of watching a movie by a user, a process of browsing a webpage by the user, and the delivered information can include but is not limited to at least one of the following: advertisement information, movie and television information. For example, in the process that a user accesses a shopping website to browse a webpage, the server may deliver advertisement information to the user, so as to recommend products in the advertisement to the user; for another example, in the process of watching a movie by a user accessing a video website, the server may release movie information to the user, so as to recommend a movie, a tv show, an art program or a video clip to the user.

The information delivery method and the information delivery device in the embodiment of the invention can be applied to a server corresponding to a website accessed by a user.

Fig. 1 is a flowchart illustrating steps of an information delivery method according to an embodiment of the present invention. As shown in fig. 1, the information delivery method may include the following steps:

step 101, a server acquires interest parameters of a target user in information categories.

When information placement is applied to recommend advertising information for a target user, the information category of the advertising information may include, but is not limited to, the industry category to which the information belongs. The industry category to which the information belongs may include, but is not limited to, at least one of: a game category, an e-commerce category, a financial category, an educational category, and the like.

When information delivery is applied to recommend movie information for a target user, the information category of the movie information may include, but is not limited to, the content category to which the information belongs. The content category to which the information belongs may include, but is not limited to, at least one of: a hedonic category, an animation category, a movie category, a television show category, an entertainment category, a health category, and so forth.

In the information release process, the server obtains interest parameters of the target user in the information category.

The server can acquire interest parameters of the target user on the information category in real time in the information delivery process. The server can also pre-acquire and store the interest parameters of each user in the information category, and in the information delivery process, the interest parameters of the target user in the information category are inquired from the pre-stored interest parameters of each user in the information category.

In the case of obtaining and storing interest parameters of each user in information categories in advance, the storage manner may include but is not limited to: any one of possible corresponding storage modes such as a table storage mode, an index storage mode, a KV (key-value) storage mode, and the like. For example, for the KV storage mode, the information can be stored in KV memories such as CouchBase (CouchBase is an open source, distributed NoSQL database, which is mainly used in the field of distributed cache and data storage, and can provide fast submillimeter-level KV storage operation through a management cache) or HiKV (high-performance key-value database).

And if the interest parameters of the users in the information categories are stored in a table storage mode, recording the line number and the column number of the interest parameters of the users in the information categories for each user. Therefore, when inquiring, the interest parameters of the target user on the information category are inquired according to the line number and the column number of the interest parameters of the target user on the information category.

If the interest parameters of the users in the information category are stored in an index storage mode, establishing an index for storing the interest parameters of the users in the information category aiming at each user. Therefore, when inquiring, the interest parameters of the target user on the information category are inquired according to the index stored by the interest parameters of the target user on the information category.

And if the interest parameters of the users on the information category are stored in a KV storage mode, aiming at each user, taking the user identification of the user as a key and taking the interest parameters of the user on the information category as a value. Therefore, when inquiring, according to the user identification of the target user, inquiring the interest parameters of the target user on the information category.

The method for acquiring the interest parameters of the user in the information categories by the server may include, but is not limited to, at least one of the following: model prediction, questionnaires, historical behavior data analysis.

In the model prediction mode, a prediction model is trained in advance, and the server predicts and obtains interest parameters of the user on the information category based on the prediction model. Details will be described later.

In the questionnaire mode, the server can issue questionnaires for users to fill in, information category options can be provided in the questionnaires for the users to select interested information categories and/or uninteresting information categories, and the server obtains interest parameters of the users on the information categories based on data in the questionnaires filled in by the users.

In the historical behavior data analysis mode, the server may obtain historical behavior data of the user, and the historical behavior data may include, but is not limited to, at least one of the following: purchase data, click data, play data, etc. The server obtains interest parameters of the user on the information category based on the historical behavior data of the user. Alternatively, the information category corresponding to the product purchased for multiple times may be used as the information category interested by the user, the information category corresponding to the movie played for multiple times may be used as the information category interested by the user, and so on.

The interest parameter of the user on the information categories is used for indicating the interest degree of the user on each information category.

In an alternative embodiment, the interest parameter of the user in the information category may be whether the user is interested in each information category.

In another alternative embodiment, the interest parameter of the user in the information category may be an interest score of the user in each information category.

In another alternative embodiment, the interest parameter of the user in the information category may be the first category set and/or the second category set. The first set of categories may include at least one first category. The second set of categories may include at least one second category or the second set of categories may be empty. The user interest level of any one of the first categories is higher than the user interest level of any one of the second categories.

And 102, the server acquires the sequencing result of the information to be released corresponding to the target user.

Information delivery may include three links: information recall, information sorting and information display. This step 102 mainly involves two links of information recall and information sorting.

In the information recalling link, the server acquires the information to be released corresponding to the target user.

Optionally, the server may adopt an experience recall manner, such as a recall manner using tag matching or the like, based on the target user, to recall the candidate information as the first information to be delivered, and to take the first information to be delivered as the information to be delivered corresponding to the target user. In this way, the order of executing the two steps, i.e., step 101 and step 102, is not limited.

Optionally, the server may recall candidate information related to the interest parameter of the target user in the information category as second information to be delivered based on the interest parameter of the target user in the information category, and use the second information to be delivered as information to be delivered corresponding to the target user. In the process of recalling the second information to be released, the existing label of each piece of information in the information base can be obtained, the existing label of the information is matched with the interest parameters of the user in the information category, and if the matching is successful, the information is related to the interest parameters of the target user in the information category.

Optionally, the first to-be-released information and the second to-be-released information recalled in the above manner may be used as the to-be-released information corresponding to the target user.

Optionally, based on the target user, the candidate information may be recalled as the first information to be delivered in an experience recall manner; and then screening out information to be released corresponding to the target user from the first information to be released based on the interest parameters of the target user in the information category. In the process of screening out the information to be released corresponding to the target user from the first information to be released, the existing label of each first information to be released can be obtained, the existing label of the first information to be released is matched with the interest parameters of the user in the information category, if the matching is successful, the first information to be released is related to the interest parameters of the target user in the information category, and therefore the first information to be released, which is successfully matched with the interest parameters of the user in the information category, is used as the information to be released corresponding to the target user.

And after the information recalling link, entering an information sequencing link. In the information sorting link, sorting the information to be released corresponding to the target user to obtain a sorting result of the information to be released.

In an alternative embodiment, the ranking model may be utilized to rank the placement information.

In another optional implementation, the information arrival rate of each to-be-delivered information corresponding to the target user may be obtained; and sequencing the information to be released based on the information arrival rate. The information arrival rate may include, but is not limited to, at least one of: click rate, conversion rate, etc. The click rate is the ratio of the number of times of information being clicked to the number of times of information being displayed, and reflects the attention degree of the information, so as to measure the attraction degree of the information to the user. The conversion rate is the ratio of the number of times of conversion behaviors (such as purchasing behaviors, downloading behaviors and the like) of the information to the number of times of clicking the information, and reflects whether the information is profitable after being released. In an implementation, the ordering may be performed according to a single information arrival rate, or according to a weighted result of multiple information arrival rates, and so on.

Of course, any other suitable sorting method may be used to sort the information to be delivered according to actual experience, which is not limited in this embodiment of the present invention.

And 103, the server adjusts the ranking result of the information to be released according to the interest parameters of the target user on the information category to obtain the adjusted ranking result.

Information delivery may include three links: information recall, information sorting and information display. The step 103 mainly relates to an information sorting link in information delivery, and mainly adjusts the sorting result obtained in the step 102. That is, in order to more accurately perform information delivery, the ranking result of the information to be delivered may be adjusted according to the interest parameters of the target user in the information category.

In an optional implementation manner, the adjusting the process of the ranking result of the information to be delivered according to the interest parameter of the target user in the information category may include: when the interest parameters of the target user on the information categories indicate that: when the category of the information to be released belongs to the information category interested by the target user, controlling the ranking of the information to be released in the ranking result to be forward, thereby increasing the display chance of the information interested by the target user; and/or, when the interest parameter of the target user in the information category indicates that: and when the category of the information to be released belongs to the information category which is not interested by the target user, controlling the ranking of the information to be released in the ranking result to be backward, thereby reducing the display chance of the information which is not interested by the user.

According to the interest parameters of the target user on the information categories, whether the categories of the information to be delivered belong to the information categories which are interested by the target user or not can be determined. The interest parameters of the user in the information categories can be divided into information categories which are interesting to the user, information categories which are less interesting to the user, information categories which may not be interesting to the user and information categories which do not have enough user data. In one possible scenario, the categories other than the categories of information of interest to the user may be unified as categories of information that are not of interest to the user.

When the interest parameter of the target user in the information category is whether the target user is interested in each information category, the information category with the interest degree of the target user being yes may be used as the information category interested by the target user, and the information category with the interest degree of the target user being no may be used as the information category not interested by the target user.

When the interest parameter of the target user in the information category is the interest score of the target user in each information category, the information category of which the interest score of the target user is greater than the interest threshold value can be used as the information category of which the target user is interested, and the information category of which the interest score of the target user is less than the non-interest threshold value can be used as the information category of which the target user is not interested. For the specific values of the interested threshold and the uninteresting threshold, any applicable value may be set according to practical experience, which is not limited in the embodiment of the present invention.

When the interest parameter of the target user in the information category is the first category set, the first category may be an information category in which the target user is interested, and other categories except the first category may be information categories in which the target user is not interested.

In the case that the interest parameter of the target user in the information category is the second category set, the second category may be an information category that is not interested by the target user, and other categories except the second category may be information categories that are interested by the target user.

In the case that the interest parameters of the target user in the information categories are the first category set and the second category set, the first category may be used as the information category in which the target user is interested, and the second category may be used as the information category in which the target user is not interested.

Optionally, in the ranking result of the information to be placed, each information to be placed corresponds to one ranking score, and the ranking score may be regarded as a priority of the information to be placed in the ranking result. Wherein the higher the priority, the earlier the ranking.

Therefore, for the case of controlling the information to be released to be ranked at the top in the ranking result, the priority of the information to be released in the ranking result can be increased.

In an alternative embodiment, the corresponding first weight coefficient when the priority is increased may be set in advance. The first weighting factor may be any suitable value greater than 1. The ranking score of the information to be launched is multiplied by the first weight coefficient, so that the ranking score of the information to be launched can be increased, the priority of the information to be launched in the ranking result is improved, and the ranking of the information to be launched in the ranking result is controlled to be advanced.

In another alternative embodiment, the corresponding first adjustment parameter when the priority is increased may be preset. The first adjustment parameter may be any suitable value greater than 0. The ranking score of the information to be released can be increased by adding the first adjusting parameter to the ranking score of the information to be released, so that the priority of the information to be released in the ranking result is improved, and the ranking of the information to be released in the ranking result is controlled to be advanced.

For the situation that the information to be released is controlled to be ranked behind the ranking result, the ranking result can be used for reducing the priority of the information to be released in the ranking result.

In an alternative embodiment, the corresponding second weight coefficient when the priority is lowered may be set in advance. The second weighting factor may be any suitable value greater than 0 and less than 1. The ranking score of the information to be launched is multiplied by the second weight coefficient, so that the ranking score of the information to be launched can be reduced, the priority of the information to be launched in the ranking result is reduced, and the ranking of the information to be launched in the ranking result is controlled to be backward.

In another alternative embodiment, the corresponding second adjustment parameter for lowering the priority may be preset. The second adjusting parameter can be any applicable value larger than 0, and the ranking score of the information to be released can be reduced by subtracting the second adjusting parameter from the ranking score of the information to be released, so that the priority of the information to be released in the ranking result is reduced, and the ranking of the information to be released in the ranking result is controlled to be backward. The second adjusting parameter can also be any applicable value smaller than 0, and the ranking score of the information to be released can be reduced by adding the second adjusting parameter to the ranking score of the information to be released, so that the priority of the information to be released in the ranking result is reduced, and the ranking of the information to be released in the ranking result is controlled to be backward.

It should be noted that, under the condition that the interest degree is the interest score, if the ranking is controlled to be forward and/or backward, the degrees of the forward ranking and/or the backward ranking may be different for the information to be released with different interest scores; if the priority is controlled to be improved and/or reduced, the degree of improving the priority and/or reducing the priority can be different aiming at the information to be released with different interest scores; if the first weight coefficient and/or the second weight coefficient are/is set, the set first weight coefficient and/or second weight coefficient can be different aiming at the information to be released with different interest scores; if the first adjustment parameter and/or the second adjustment parameter are/is set, the set first adjustment parameter and/or second adjustment parameter may be different for the information to be released with different interestingness scores.

And 104, the server puts information into the target user according to the adjusted sorting result.

In an optional implementation manner, the server filters the information to be delivered sorted at the top as target delivery information from the adjusted sorting result, and delivers the target delivery information to the target user. And the video website displays the target release information to the target user and successfully completes the information release.

And as for the mode of recalling the first information to be released by adopting an experience recall mode and taking the first information to be released as the information to be released corresponding to the target user, the mode can recall the first information to be released more simply, conveniently and quickly. In the process of adjusting the ranking result of the information to be released, the ranking result is adjusted based on the interest parameters of the target user on the information category, so that the adjusted ranking result is inclined to the information which the user is interested in.

And for recalling second information to be released related to the interest parameters of the target user in the information category, taking the second information to be released as the information to be released corresponding to the target user, wherein the recalled second information to be released is related to the interest parameters of the target user in the information category, so that the second information to be released can be more matched with the requirements of the user. In the process of adjusting the ranking result of the information to be released, the interest parameters of the target user are considered again for adjustment, so that the relevance between the adjusted ranking result and the interest parameters of the user on the information category is larger, and the subsequent information releasing result is more accurate.

For the mode that the first information to be released and the second information to be released which are recalled according to the mode are taken as the information to be released corresponding to the target user, the two kinds of information to be released are recalled according to the mode, the recalled information to be released can be more accurate and more comprehensive, and therefore the information in the follow-up sequencing result is more comprehensive.

The method comprises the steps of firstly recalling first information to be released by adopting an experience recall mode, then screening information to be released corresponding to a target user from the first information to be released based on interest parameters of the target user in an information category, and the method can recall the first information to be released simply and quickly, and then screen information to be released which is more matched with requirements of the user from the first information to be released by utilizing the interest parameters of the target user in the information category.

In the embodiment of the invention, the information is released by combining the interest parameters of the single user on the information category, and the sequencing result of the information to be released can be adjusted based on the information category, so that the information released by any single user can be more inclined to the related information of the interest category of the single user to a certain extent, the accuracy of information release is improved, the information is released to people who really have needs, the experience of the audience is optimized, the disturbance to non-target people is reduced, and the information release effect is enhanced.

As described in step 101, in the embodiment of the present invention, the interest parameter of the user in the information category may be obtained through various means, including but not limited to, through a model prediction manner. In this way, in order to predict the interest parameters of the user in the information category more quickly and easily, a prediction model for predicting the interest parameters of the user in the information category may be set up in advance.

Next, a process of building the prediction model will be described.

In an optional implementation manner, for each information category, a prediction model corresponding to the information category may be generated, that is, one information category may correspond to one prediction model, and the prediction model is used to predict the degree of interest of the user in the information category corresponding to the prediction model.

In another alternative embodiment, a prediction model for predicting the respective interest level of the user in the various information categories may be generated uniformly for the various information categories.

In another alternative embodiment, for each user, a prediction model corresponding to the user may also be generated. The prediction model corresponding to the user may be a prediction model corresponding to one user, and the prediction model is used for predicting the interest degree of the user in each information category.

As shown in fig. 2, the prediction model building process may include the following steps:

step 201, sample marking and feature extraction are performed on user data.

User data for a large number of users is obtained from the internet. User data may include, but is not limited to, network behavior data, information behavior data, etc. of the user. The network behavior data of the user may include, but is not limited to, at least one of the following: gender, age, region, access time, access duration and other data of the user. The information behavior data of the user may include, but is not limited to, at least one of: the data of the user on the behavior of the information (such as playing, clicking, converting and other behaviors), the time of the behavior, the information category of the behavior and the like.

And cleaning the user data. During the data cleansing process, invalid data may be deleted from the user data, and the invalid data may include, but is not limited to, at least one of: duplicate data, anomalous data (e.g., age 150, etc.), etc., thereby preserving valid data. During the cleaning process of the abnormal data, the abnormal data in the user data can be obtained by performing abnormal analysis on the user data. The anomaly analysis means may include, but is not limited to, at least one of: an anomaly analysis model, a threshold value judgment and the like. Optionally, the retained valid data may be further normalized to obtain valid data with a regular data format, for example, normalizing the age to a digital format, normalizing the time to a year/month/day/hour/minute/second format, and the like. And after the data is cleaned, integrating the normalized effective data based on the Identification (ID) of the user to obtain the effective data corresponding to each user.

For any one user, the sample is manually marked based on the information behavior data of the user. In implementation, the times of the user performing a conversion action, a click action, a play action, and the like on information of the same information category may be counted. If the user generates conversion behavior, click behavior, play behavior and the like exceeding the preset times on the information of a certain information category, marking that the user is interested in the information category; if the information of a certain information category is exposed for multiple times but the user does not have conversion, clicking behavior, playing behavior and the like, the information category is marked as uninterested in the user. For the preset number, any suitable value may be set according to practical experience, for example, 70%, 80%, etc. of the information exposure number may be set as the preset number.

For any user, extracting user attribute characteristic data from the network behavior data of the user, and extracting behavior characteristic data of the user on the information category from the information behavior data of the user. And taking the attribute characteristic data of the user and/or the behavior characteristic data of the user on the information category as the characteristic data of the user.

The user's own attribute feature data may include, but is not limited to, at least one of: gender, age, location, etc. of the user. The users have different attribute characteristic data, and the users have different interest degrees in the information categories. For example, if the gender of the user is male, the user may be interested to a high degree in a game category, an animation category, etc., and less in an education category, an entertainment category, etc.; if the gender of the user is female, the user may be more interested in education categories, entertainment categories, etc. and less interested in game categories, entertainment categories, etc.; if the user's age belongs to the teenage age group, the user may have a high level of interest in education categories, animation categories, etc., and a low level of interest in finance categories, health preserving categories, etc.; if the user's age belongs to the elderly age group, the user may be interested to a high degree in a financial category, a health category, etc., and may be interested to a low degree in an education category, an animation category, etc. Therefore, the user self attribute feature data can represent the interest degree of the user in the information category.

The behavior feature data of the user on the information category may include, but is not limited to, at least one of the following: the type of information that a play action occurs, the type of information that a click action occurs, the type of information that a conversion action occurs, and so on. The behavior characteristic data of the user on the information category can represent the interest degree of the user on the information category. For example, if a user performs a multi-play action, a click action, a conversion action, etc. on information of an information category, it indicates that the user has a high interest level in the information category; if the user does not perform playing behavior, clicking behavior, converting behavior and the like on the information of a certain information category, the user is indicated to have low interest degree on the information category.

Step 202, generating sample data by combining the sample marks and the data obtained by feature extraction.

After the sample marking and the feature extraction are respectively carried out on each user, each user can be used as a sample user, the sample data is formed by combining the sample marking and the data obtained by the feature extraction, and the sample data is randomly divided into a training set and a testing set according to a preset proportion. For the specific numerical value of the preset ratio, any suitable numerical value may be set according to practical experience, for example, the ratio of the training sample data in the training set to the test sample data in the test set is 8:2, 7:3, and the like, which is not limited in this embodiment of the present invention.

It should be understood that the sample data generated here may differ based on the model being built. For example, if a prediction model is generated for an information category, for any information category, one sample data may include feature data of a sample user and label information of the sample user, where the label information of the sample user is used to indicate a degree of interest of the sample user in the any information category. If a manner of uniformly generating a prediction model for various information categories is adopted, one sample data may include feature data of a sample user and label information of the sample user, where the label information of the sample user is used to indicate the degree of interest of the sample user in each information category. For example, a flag information of "1" indicates interest, a flag information of "0" indicates not interest, and so on.

And step 203, training a preset basic model by using the training set.

The base model is a model with a classification function that has not been trained. If a predictive model is generated for an information category, the base model may be a binary model that has not been trained. The classification task of the two-class model has two classes, each sample belongs to one of the two classes, and the label is 0 or 1. If a unified predictive model generation is used for various information categories, the base model may be a multi-label classification model that has not been trained. The classification task of the multi-label classification model has n classes, each sample belongs to a plurality of the n classes, and each sample has a plurality of labels.

If a predictive model is generated for an information category, then one information category corresponds to one base model. Aiming at any information category, in the process of training the basic model corresponding to the information category by using a training set, the input of the basic model corresponding to the information category is the characteristic data of a training sample user contained in training sample data, and the output is the predicted interest degree of the training sample user on the information category. If a mode of uniformly generating a prediction model aiming at various information categories is adopted, in the process of training the basic model by using the training set, the input of the basic model is the characteristic data of the training sample user contained in the training sample data, and the output is the prediction interest degree of the training sample user on various information categories respectively.

In an alternative embodiment, the loss value is calculated based on the predicted interest level of the training sample user in the information category and the label information of the training sample user. The loss value can represent the deviation degree of the prediction result and the mark information, and the smaller the loss value is, the better the robustness of the model is. Therefore, it can be determined that the model preliminary training is completed when the loss value is less than the preset loss threshold. For the specific value of the loss threshold, any suitable value may be selected by those skilled in the art according to practical experience, and may be set to 0.1, 0.2, 0.3, and so on. In another alternative embodiment, it may be set that the preliminary training of the model is determined to be completed when a preset number of iterations is reached. For the specific number of iterations, any suitable value may be selected by one skilled in the art based on practical experience.

In the embodiment of the invention, any model structure can be selected as the prediction model. May include, but is not limited to: an LR (Logistic Regression) model, an SVM (Support Vector Machine) model, an FM (Factorization Machine) model, a DNN (Deep Neural Networks) model, and the like.

Considering that the prediction model predicts the interest degree of the user in the information category according to the feature data of the user, the combination of different feature data can more accurately indicate the interest degree of the user in the information category compared with the feature data alone. For example, for the individual characteristic data that the gender of the user is male and the individual characteristic data that the information category of the conversion behavior of the user is the game category, it can be found from practical experience that the male may prefer the information of the game category, and thus the two individual characteristic data of the male and the game category have a larger correlation. Therefore, the sex of the user can be used as male and the information category of the conversion behavior of the user can be used as game category, and the two separate feature data are combined to form a new feature data, namely a male + game category. Therefore, in order to combine different features with relevance to form a new feature so as to enrich the relevance between the features, in the embodiment of the present invention, an FM model capable of autonomously learning a hidden relationship between the features may be selected to train a prediction model on a training set, thereby achieving the above-mentioned purpose. The FM model has the greatest advantage over other traditional machine learning models in that the second-order combination of the features is considered, namely, the features can be combined pairwise to form a new feature, the relevance between the features can be fully learned, and the generalization capability of the algorithm model is enhanced.

And 204, evaluating the training process of the basic model by using the test set and/or the off-line data to obtain the prediction model which passes evaluation.

And evaluating the preliminarily trained prediction model by using the test set and/or the off-line data.

The mode of evaluating by independently utilizing the test set can directly utilize the partial sample data as the test set to evaluate, and does not need to independently carry out off-line investigation and collect off-line data, so that the processing process is simpler and more convenient. The evaluation mode of independently utilizing the offline data can obtain more accurate and real offline data through offline investigation, thereby improving the accuracy of evaluation.

By using the combined evaluation mode of test set evaluation and off-line data evaluation, considering that in the mode based on test set evaluation alone, because the information category interested by the user is marked based on the historical behavior of the user, the situation that the marking result is not really interested by the user may exist, the error brought by the test set can be made up by using the more accurate data of off-line data, and the accuracy of the prediction model is further ensured.

The following are described separately.

In one aspect, the process of evaluating the training process of the base model using the test set may include: inputting characteristic data of a test sample user contained in test sample data into the prediction model to obtain a second prediction result output by the prediction model, wherein the second prediction result is used for indicating the prediction interest degree of the test sample user on the information category; calculating a first evaluation index of the prediction model based on the second prediction result and the actual interest degree of the test sample user in the information category contained in the test sample data; and determining that the evaluation is passed under the condition that the first evaluation index meets a first preset condition.

The first assessment indicator may include, but is not limited to, at least one of an accuracy rate or an AUC (Area Under Curve).

The method includes the steps of obtaining a first total number of samples predicted to be interesting and actually interesting, and obtaining a second total number of samples predicted to be interesting and actually interesting and predicted to be interesting and not actually interesting, wherein the precision rate is a ratio obtained by dividing the first total number by the second total number. The accuracy rate can reflect the ability of the prediction model to correctly predict the sample actually interested as being interested, so that the higher the accuracy rate is, the stronger the ability of the prediction model to correctly predict the interest is, and therefore the prediction of the information category of interest can be biased more.

ROC (Receiver Operating Characteristic Curve) is a comprehensive index reflecting continuous variables of sensitivity and specificity, and each point on the ROC Curve reflects the sensitivity to the same signal stimulus. The horizontal axis of the ROC curve represents the false positive class rate, namely the proportion of the samples predicted to be interesting but not actually interesting to all the samples not actually interesting; the vertical axis represents true class rate, i.e., the proportion of samples predicted to be of interest and actually of interest to all samples actually of interest. And after obtaining the ROC curve corresponding to the prediction model, calculating the area under the ROC curve, wherein the area under the ROC curve is the AUC. The AUC value is a probability value, which means the probability that the prediction model ranks the degree of interest of a sample marked as of interest before a sample marked as not of interest, and the greater the AUC value, the more likely the prediction model will rank the degree of interest of a sample marked as of interest before a sample marked as not of interest, thereby enabling better classification. Thus a higher AUC indicates a better discriminative power of the predictive model.

In a case where the first evaluation index includes the precision ratio, the first preset condition may be that the precision ratio is greater than a first preset threshold.

In the case where the first evaluation index includes AUC, the first preset condition may be that AUC is greater than the second preset threshold.

In the case that the first evaluation index includes the accuracy rate and the AUC, the first preset condition may be that the accuracy rate is greater than a first preset threshold, and/or the AUC is greater than a second preset threshold.

For the specific values of the first preset threshold and the second preset threshold, any applicable value may be selected according to practical experience, which is not limited in the embodiment of the present invention. For example, the first preset threshold may be 0.8, 0.85, etc., and the second preset threshold may be 0.8, 0.85, etc.

On the other hand, when the training process of the basic model is evaluated by using the offline data, the offline data of the offline user needs to be collected. The offline user may be any user of the website, and optionally, the offline user may be a part of the sample users, or may be another user other than the sample users. The server may collect offline data of the offline user by way of offline research. In the implementation, the server may issue a questionnaire for offline user to fill in, the questionnaire may provide input items of user's own attribute data such as gender, age, region, and the like, and information category options, the offline user may select an interested information category and/or an uninteresting information category, and the server obtains offline data of the offline user based on data in the questionnaire filled in by the user.

If a prediction model is generated for an information category, for any information category, an offline data may include feature data of an offline user and label information of the offline user, where the label information of the offline user is used to indicate a degree of interest of the offline user in the any information category.

If a manner of uniformly generating a prediction model for various information categories is adopted, one piece of offline data may include feature data of an offline user and tag information of the offline user, where the tag information of the offline user is used to indicate the degree of interest of the offline user in each information category. The feature data of the offline user may include, but is not limited to, at least one of the following: attribute characteristic data of the offline user and behavior characteristic data of the offline user on the information category.

The process of evaluating the training process of the base model using the offline data may include: inputting feature data of an offline user contained in the offline data into the prediction model to obtain a third prediction result output by the prediction model, wherein the third prediction result is used for indicating the prediction interest degree of the offline user on information categories; calculating a second evaluation index of the prediction model based on the third prediction result and the actual interest degree of the offline user in the information category contained in the offline data; and determining that the evaluation is passed under the condition that the second evaluation index meets a second preset condition.

The second evaluation index may include, but is not limited to, accuracy. Accuracy refers to the ratio of the number of samples that are all predicted to be correct divided by the total number of samples. The accuracy can reflect the correct prediction capability of the prediction model, so that the higher the accuracy is, the higher the possibility that the prediction model is correct in prediction is, and the accuracy of the prediction model can be intuitively reflected through the accuracy. In the case where the second evaluation index includes an accuracy rate, the second preset condition may be that the accuracy rate is greater than a third preset threshold. For the specific value of the third preset threshold, any suitable value may be selected according to practical experience, which is not limited in the embodiment of the present invention. For example, the third preset threshold may be 0.8, 0.85, etc.

For any user, the process of predicting the interest parameters of the any user in the information categories by using the prediction model may include: acquiring feature data of any user aiming at any user; inputting the feature data of any user into a pre-generated prediction model to obtain a fourth prediction result output by the prediction model, wherein the fourth prediction result is used for indicating the interest degree of any user in the information category; and determining the interest parameter of the any user in the information category based on the interest degree of the any user in the information category.

After the degree of interest of any user in each information category is predicted, an interest parameter of any user in each information category can be obtained, and the interest parameter is used for indicating the degree of interest of any user in each information category.

In an alternative embodiment, whether each information category is interested by the user or not can be used as the interest parameter of the user on the information category.

In another alternative embodiment, the interest score of the user in each information category can be used as the interest parameter of the user in the information category.

In another alternative embodiment, the first category set and/or the second category set may be divided, and the first category set and/or the second category set may be used as the interest parameter of the user in the information category.

For the case where only the first set of categories is partitioned, the first set of categories may include at least one first category. In the case where the degree of interest indicates whether the information category is of interest to the user, the information category of which the degree of interest is yes may be taken as the first category; in the case where the interest degree indicates the interest score of the user for the information category, an interest threshold may be set in advance, and the information category whose interest score is greater than the interest threshold may be taken as the first category.

For the case where only the second set of categories is divided, the second set of categories may include at least one second category, or the second set of categories may be empty. In the case where the degree of interest indicates whether the information category is of interest to the user, the information category of which the degree of interest is no may be taken as the second category; in the case where the interest degree indicates the interest score of the user for the information category, a non-interest threshold may be set in advance, and the information category whose interest score is smaller than the non-interest threshold may be taken as the second category.

For the case of dividing the first set of categories and the second set of categories, the first set of categories may include at least one first category, the second set of categories may include at least one second category, or the second set of categories may be empty.

In the event that the level of interest indicates whether the information category is of interest to the user, in one implementation, the information category of which the level of interest is yes may be taken as a first category and the information category of which the level of interest is no may be taken as a second category; in another implementation, the information category with the interest degree of yes may be taken as a first category, and the information categories other than the first category may be taken as a second category; in another implementation, the information category with the negative degree of interest may be taken as the second category, and the information categories other than the second category may be taken as the first category.

In the case where the interest degree indicates the user's interest score for the information category, an interest threshold and a non-interest threshold may be set in advance. In one implementation, the information category with the interest score greater than the interest threshold may be taken as a first category, and the information category with the interest score less than the non-interest threshold may be taken as a second category; in another implementation, an information category with an interest score greater than an interest threshold may be taken as a first category, and an information category other than the first category may be taken as a second category; in another implementation, the information category with the interest score smaller than the non-interest threshold may be taken as the second category, and the information category other than the second category may be taken as the first category.

In the embodiment of the invention, on the first hand, the construction of the prediction model is combined with the characteristic data of the user, so that the interest of the user in different types of information can be more effectively mined; in the second aspect, the prediction model evaluation method comprises two parts, namely model index evaluation based on the accuracy of the test set, such as AUC (AUC), and accuracy evaluation based on offline data of an offline user, so that the accuracy of the prediction model can be effectively ensured.

As shown in fig. 3, the information delivery apparatus may include the following modules:

a first obtaining module 301, configured to obtain an interest parameter of a target user in an information category;

a second obtaining module 302, configured to obtain a ranking result of the information to be released corresponding to the target user;

the adjusting module 303 is configured to adjust the ranking result of the information to be released according to the interest parameter of the target user in the information category, so as to obtain an adjusted ranking result;

and a releasing module 304, configured to release information to the target user according to the adjusted sorting result.

As shown in fig. 4, the information delivery apparatus may include the following modules:

a first obtaining module 401, configured to obtain an interest parameter of a target user in an information category;

a second obtaining module 402, configured to obtain a ranking result of the information to be released corresponding to the target user;

an adjusting module 403, configured to adjust a ranking result of the information to be delivered according to the interest parameter of the target user in the information category, to obtain an adjusted ranking result;

and a releasing module 404, configured to release information to the target user according to the adjusted sorting result.

Optionally, the information category includes at least one of: the industry category to which the information belongs, and the content category to which the information belongs.

Optionally, the interest parameter of the target user in the information category is used to indicate the interest degree of the target user in each information category; the adjusting module 403 includes: a first adjusting unit, configured to, when the interest parameter of the target user in the information category indicates: when the category of the information to be released belongs to the information category which is interested by the target user, controlling the ranking of the information to be released in the ranking result to be advanced; and/or, a second adjusting unit, configured to, when the interest parameter of the target user in the information category indicates: and when the category of the information to be released belongs to the information category which is not interested by the target user, controlling the ranking of the information to be released in the ranking result to be backward.

Optionally, the first obtaining module 401 includes: the characteristic acquisition unit is used for acquiring characteristic data of the target user, wherein the characteristic data comprises user attribute characteristic data and/or behavior characteristic data of the user on information categories; the model prediction unit is used for inputting the characteristic data of the target user into a pre-generated prediction model to obtain a first prediction result output by the prediction model, and the first prediction result is used for indicating the interest degree of the target user in the information category; and the parameter acquisition unit is used for determining the interest parameters of the target user on the information categories based on the interest degree of the target user on the information categories.

Optionally, the first obtaining module 401 includes: the query unit is used for querying the interest parameters of the target user on the information category from the pre-stored interest parameters of the users on the information category; the interest parameters of any user in the information category are obtained by prediction based on the characteristic data of any user by using a pre-generated prediction model.

Optionally, the prediction model is a factorizer FM model.

Optionally, the apparatus further comprises: a third obtaining module 405, configured to obtain sample data, where the sample data includes a training set and a test set; a training module 406, configured to train a preset basic model by using the training set, where an input of the basic model is feature data of a training sample user included in training sample data, and an output is a predicted interest degree of the training sample user in an information category; and the evaluation module 407 is configured to evaluate the training process of the basic model by using the test set and/or the offline data, so as to obtain the prediction model that passes the evaluation.

Optionally, the evaluation module 407 comprises: the first prediction unit is used for inputting the characteristic data of the test sample user contained in the test sample data into the prediction model to obtain a second prediction result output by the prediction model, and the second prediction result is used for indicating the prediction interest degree of the test sample user on the information category; a first calculating unit, configured to calculate a first evaluation index of the prediction model based on the second prediction result and an actual interest level of the test sample user in an information category included in the test sample data; the first assessment indicator comprises an accuracy rate and/or an area under the curve AUC; and the first determination unit is used for determining that the evaluation is passed under the condition that the first evaluation index meets a first preset condition.

Optionally, the evaluation module 407 comprises: the second prediction unit is used for inputting the characteristic data of the offline user contained in the offline data into the prediction model to obtain a third prediction result output by the prediction model, and the third prediction result is used for indicating the degree of the offline user interested in the prediction of the information category; a second calculation unit, configured to calculate a second evaluation index of the prediction model based on the third prediction result and an actual interest level of the offline user in an information category included in the offline data; the second assessment indicator comprises an accuracy rate; and the second determination unit is used for determining that the evaluation is passed under the condition that the second evaluation index meets a second preset condition.

Optionally, the second obtaining module 402 includes: a first information obtaining unit, configured to obtain information to be delivered corresponding to the target user, where the information to be delivered includes: the system comprises first information to be released and/or second information to be released, wherein the first information to be released is candidate information recalled based on the target user, and the second information to be released is candidate information determined based on interest parameters of the target user in information categories; and the first sequencing unit is used for sequencing the information to be released corresponding to the target user to obtain a sequencing result of the information to be released.

Optionally, the second obtaining module 402 includes: a second information obtaining unit, configured to obtain first information to be delivered corresponding to the target user, where the first information to be delivered is candidate information recalled based on the target user; the information screening unit is used for screening the information to be released corresponding to the target user from the first information to be released based on the interest parameters of the target user in the information category; and the second sorting unit is used for sorting the information to be launched corresponding to the target user to obtain a sorting result of the information to be launched.

Optionally, the first sequencing unit is specifically configured to obtain an information arrival rate of each to-be-delivered information corresponding to the target user, where the information arrival rate includes a click rate and/or a conversion rate; and sequencing the information to be released based on the information arrival rate.

In the embodiment of the invention, information is released by combining the interest parameters of a single user on the information category, and the sequencing result of the information to be released can be adjusted based on the information category, so that the released information of any single user can be more inclined to the related information of the interest category of the single user to a certain extent, and the relevance between the released information and the user is favorably improved, thereby improving the accuracy of information release, releasing the information to people really having needs, optimizing the experience of the audience, reducing the disturbance to non-target people and enhancing the effect of information release.

An embodiment of the present invention further provides an electronic device, as shown in fig. 5, including a processor 501, a communication interface 502, a memory 503 and a communication bus 504, where the processor 501, the communication interface 502 and the memory 503 complete mutual communication through the communication bus 504.

A memory 503 for storing a computer program;

the processor 501 is configured to implement the information delivery method according to any of the above embodiments when executing the program stored in the memory 503.

The communication bus mentioned in the above terminal may be a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, or the like. The communication bus may be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, only one thick line is shown, but this does not mean that there is only one bus or one type of bus.

The communication interface is used for communication between the terminal and other equipment.

The Memory may include a Random Access Memory (RAM) or a non-volatile Memory (non-volatile Memory), such as at least one disk Memory. Optionally, the memory may also be at least one memory device located remotely from the processor.

The Processor may be a general-purpose Processor, and includes a Central Processing Unit (CPU), a Network Processor (NP), and the like; the Integrated Circuit may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, a discrete Gate or transistor logic device, or a discrete hardware component.

In another embodiment of the present invention, a computer-readable storage medium is further provided, where instructions are stored in the computer-readable storage medium, and when the instructions are executed on a computer, the computer is enabled to implement the information delivery method according to any one of the above embodiments.

In another embodiment of the present invention, a computer program product containing instructions is further provided, which when run on a computer, causes the computer to implement the information delivery method according to any one of the above embodiments.

In the above embodiments, the implementation may be wholly or partially realized by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loaded and executed on a computer, cause the processes or functions described in accordance with the embodiments of the invention to occur, in whole or in part. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable device. The computer instructions may be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another, for example, from one website site, computer, server, or data center to another website site, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device, such as a server, a data center, etc., that incorporates one or more of the available media. The usable medium may be a magnetic medium (e.g., floppy Disk, hard Disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., Solid State Disk (SSD)), among others.

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

All the embodiments in the present specification are described in a related manner, and the same and similar parts among the embodiments may be referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the system embodiment, since it is substantially similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment.

The above description is only for the preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention shall fall within the protection scope of the present invention.

Claims

1. An information delivery method, comprising:

obtaining interest parameters of a target user on information categories;

2. The method of claim 1, wherein the categories of information include at least one of: the industry category to which the information belongs, and the content category to which the information belongs.

3. The method according to claim 1, wherein the interest parameter of the target user in the information categories is used for indicating the interest degree of the target user in each information category;

the adjusting the ranking result of the information to be released according to the interest parameters of the target user in the information category includes:

when the interest parameters of the target user on the information categories indicate that: when the category of the information to be released belongs to the information category which is interested by the target user, controlling the ranking of the information to be released in the ranking result to be advanced;

and/or the presence of a gas in the gas,

when the interest parameters of the target user on the information categories indicate that: and when the category of the information to be released belongs to the information category which is not interested by the target user, controlling the ranking of the information to be released in the ranking result to be backward.

4. The method of claim 1, wherein the obtaining of the interest parameters of the target user in the information category comprises:

acquiring characteristic data of the target user; the characteristic data comprises attribute characteristic data of the user and/or behavior characteristic data of the user on the information category;

inputting the characteristic data of the target user into a pre-generated prediction model to obtain a first prediction result output by the prediction model, wherein the first prediction result is used for indicating the interest degree of the target user in the information category;

and determining the interest parameters of the target user on the information categories based on the interest degree of the target user on the information categories.

5. The method of claim 1, wherein the obtaining of the interest parameters of the target user in the information category comprises:

inquiring the interest parameters of the target user on the information category from the pre-stored interest parameters of the users on the information category;

the interest parameters of any user in the information category are obtained by prediction based on the characteristic data of the user by using a pre-generated prediction model.

6. The method according to claim 4 or 5, characterized in that the predictive model is a factorizer FM model.

7. The method according to claim 4 or 5, characterized in that the method further comprises:

obtaining sample data, wherein the sample data comprises a training set and a test set;

training a preset basic model by using the training set, wherein the input of the basic model is the characteristic data of a training sample user contained in training sample data, and the output is the predicted interest degree of the training sample user on information categories;

and evaluating the training process of the basic model by using the test set and/or the off-line data to obtain the prediction model which passes evaluation.

8. The method of claim 7, wherein evaluating the training process of the base model using the test set comprises:

inputting characteristic data of a test sample user contained in test sample data into the prediction model to obtain a second prediction result output by the prediction model, wherein the second prediction result is used for indicating the prediction interest degree of the test sample user on the information category;

calculating a first evaluation index of the prediction model based on the second prediction result and the actual interest degree of the test sample user in the information category contained in the test sample data; the first assessment indicator comprises an accuracy rate and/or an area under the curve AUC;

and determining that the evaluation is passed under the condition that the first evaluation index meets a first preset condition.

9. The method of claim 7, wherein evaluating the training process of the base model using offline data comprises:

inputting feature data of an offline user contained in the offline data into the prediction model to obtain a third prediction result output by the prediction model, wherein the third prediction result is used for indicating the prediction interest degree of the offline user on information categories;

calculating a second evaluation index of the prediction model based on the third prediction result and the actual interest degree of the offline user in the information category contained in the offline data; the second assessment indicator comprises an accuracy rate;

and determining that the evaluation is passed under the condition that the second evaluation index meets a second preset condition.

10. The method according to claim 1, wherein the obtaining of the ranking result of the information to be delivered corresponding to the target user comprises:

obtaining information to be released corresponding to the target user, wherein the information to be released comprises: the system comprises first information to be released and/or second information to be released, wherein the first information to be released is candidate information recalled based on the target user, and the second information to be released is candidate information determined based on interest parameters of the target user in information categories;

and sequencing the information to be launched corresponding to the target user to obtain a sequencing result of the information to be launched.

11. The method according to claim 1, wherein the obtaining of the ranking result of the information to be delivered corresponding to the target user comprises:

acquiring first information to be launched corresponding to the target user, wherein the first information to be launched is based on candidate information recalled by the target user;

screening out information to be released corresponding to the target user from the first information to be released based on the interest parameters of the target user in the information category;

12. The method according to claim 10 or 11, wherein the sorting the information to be delivered corresponding to the target user comprises:

acquiring the information arrival rate of each to-be-released information corresponding to the target user, wherein the information arrival rate comprises click rate and/or conversion rate;

and sequencing the information to be released based on the information arrival rate.

13. An information delivery apparatus, comprising:

14. An electronic device is characterized by comprising a processor, a communication interface, a memory and a communication bus, wherein the processor and the communication interface are used for realizing mutual communication by the memory through the communication bus;

a memory for storing a computer program;

a processor for implementing the method of any one of claims 1 to 12 when executing a program stored in the memory.

15. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the method according to any one of claims 1 to 12.