CN111222030A

CN111222030A - Information recommendation method and device and electronic equipment

Info

Publication number: CN111222030A
Application number: CN201811427772.8A
Authority: CN
Inventors: 潘岸腾
Original assignee: Alibaba Group Holding Ltd
Current assignee: Alibaba Group Holding Ltd
Priority date: 2018-11-27
Filing date: 2018-11-27
Publication date: 2020-06-02
Anticipated expiration: 2038-11-27
Also published as: CN111222030B

Abstract

The invention discloses an information recommendation method and device and electronic equipment. The method comprises the following steps: acquiring a plurality of pieces of candidate information according to historical behavior data related to the information acquisition historical behaviors; based on the information recommendation model, respectively acquiring the information recommendation score of each piece of candidate information to the target user according to the information portrait of each piece of candidate information and the user portrait of the target user; and selecting candidate information of which the information recommendation score meets the preset recommendation condition as target information to be recommended to the target user.

Description

Information recommendation method and device and electronic equipment

Technical Field

The present invention relates to the field of information technologies, and in particular, to an information recommendation method and apparatus, and an electronic device.

Background

With the rapid development of the mobile internet technology and the popularization of terminal intelligence, accessing network browsing information through electronic devices such as mobile phones, tablet computers, notebook computers and the like has become an important way for users to acquire information in daily life.

With the explosive increase of the amount of information available on the network, how to quickly acquire information meeting the self-demand from massive information is the most concerned when users browse information. Accordingly, Feed stream products (e.g., push short video, news headline, etc. information applications) that provide fragmented information (e.g., short video, pictures, news, etc.) in the form of Feed streams for quick browsing by users are highly appreciated by users.

However, when the Feed stream product provides fragmented information to a user in a Feed stream form, information with a high click rate is usually recommended to the user according to the click rate of the information, so that the preference of the user for the information can only be measured by the click behavior of the user for the information, the preference degree of the user for the information cannot be accurately and effectively obtained, and the requirement of the user for obtaining the information cannot be actually met.

Disclosure of Invention

An object of the present invention is to provide a new technical solution for recommending information.

According to a first aspect of the present invention, there is provided an information recommendation method, including:

acquiring a plurality of pieces of candidate information according to historical behavior data related to the information acquisition historical behaviors;

based on an information recommendation model, respectively acquiring information recommendation scores of each piece of candidate information to a target user according to an information portrait of each piece of candidate information and a user portrait of the target user;

and selecting the candidate information of which the information recommendation score meets the preset recommendation condition as target information to be recommended to the target user.

Optionally, the historical behavior data at least includes one of a historical information record, an information source record and an overall information acquisition record; the historical information record comprises information identification of the historical information which is completely acquired by the target user; the information source record comprises a source identifier of an information source concerned by the target user; the whole information acquisition record comprises the time when each piece of information is acquired by the user in the information database;

the step of acquiring a plurality of pieces of candidate information includes:

when the historical behavior data comprises the historical information record, acquiring an information preference score of the target user for each piece of information in an information database according to the historical information record, and selecting the information of which the descending sorting order of the information preference score meets a preset first sorting range as the candidate information;

when the historical behavior data comprises the information source record, selecting information which is issued by each information source in the information source record within a preset issuing time period and has been completely acquired, wherein the descending order of the times of the information sources conforms to a preset second ordering range, and using the information as the candidate information;

when the historical behavior data comprises an overall information acquisition record, acquiring the total times of each piece of information in the information database acquired by a user in a preset recording time period according to the overall information acquisition record, and selecting the information of which the descending order of the total times acquired by the user accords with a preset third ordering range as the candidate information.

Further optionally, when the historical behavior data includes the historical information record, the step of obtaining, according to the historical information record, an information preference score of the target user for each piece of information included in an information database includes:

summing information correlation coefficients of the information and each piece of historical information recorded by the historical information to obtain information preference scores of the information;

wherein the information correlation coefficient is determined according to a first user set corresponding to the information and a second user set corresponding to the historical information; the first user set comprises all users which have completely acquired the information; and the second user set comprises all the users which completely acquire the historical information.

Optionally, the information recommendation model includes a plurality of model features and an information recommendation weight set, each model feature has a unique feature identifier, the information recommendation weight set includes a constant weight and a plurality of feature weights, and each feature weight uniquely corresponds to one model feature;

the step of obtaining the information recommendation score of each piece of candidate information to the target user comprises the following steps:

determining a value corresponding to each model feature in the information recommendation model according to the information portrait of the candidate information and the user portrait of the target user;

wherein the information representation is representation data acquired within a preset information statistics period and related to a state in which the corresponding information is acquired by a user; the user portrait is portrait data which is acquired within a preset user statistic time period and is related to corresponding information acquisition behaviors of the user;

and acquiring the information recommendation score of the candidate information to a target user according to the information recommendation weight set and the determined value corresponding to each model feature.

In a further alternative,

the information portrait at least comprises average complete acquisition times, a forward feedback ratio, a complete acquisition ratio and an information label of the corresponding information; the average complete acquisition times are times of complete acquisition of the information in each preset time unit in the information statistics period; the forward feedback rate is the ratio of the number of times of obtaining forward feedback after the information is acquired in the information counting time period to the number of times of issuing the information; the complete acquisition ratio is a ratio of the number of times the information is completely acquired within the information statistics period and the number of times the information is issued;

the user portrait at least comprises a complete acquisition information record, a forward feedback information record and an information preference label of the corresponding user; the complete acquired information record comprises information identifications of information which corresponds to the user and completely acquires information meeting a first preset number in the user counting time period; the forward feedback information record comprises information marks of information which accords with a second preset number and is given forward feedback by the corresponding user in the user counting time period; the information preference tag is an information tag of which the repetition times in the information tags of all the information recorded in the complete acquired information record and/or the forward feedback information record meet a preset repetition threshold;

the plurality of model features included in the information recommendation model respectively belong to a plurality of different feature classes, wherein the plurality of feature classes include a first feature class which is cross-related to a complete acquisition information record of the user and an information identifier of the information, a second feature class which is cross-related to a forward feedback information record of the user and an information identifier of the information, a third feature class which is cross-related to an information preference tag of the user and an information tag of the information, a fourth feature class which is related to an average complete acquisition number of the information, a fifth feature class which is related to a forward feedback ratio of the information, and a sixth feature class which is related to the complete acquisition ratio;

the step of determining a value corresponding to each model feature included in the information recommendation model according to the information portrait of the candidate information and the user portrait of the target user comprises:

setting the value of the model feature to be 1 when determining that the model feature comprises a model feature in the information recommendation model according to the information portrait of the candidate information and the user portrait of the target user;

and setting the value of the model feature to be 0 when determining that the model feature does not have one model feature included in the information recommendation model according to the information portrait of the candidate information and the user portrait of the target user.

Optionally, the method further includes a step of training to obtain the information recommendation model, including:

according to a plurality of training samples obtained in a statistical time period, a plurality of model features included in the information recommendation model are constructed;

wherein each of said training samples comprises said user representation of a user and said information representation of an information;

respectively determining the value of each model characteristic of each training sample;

respectively determining an information recommendation sub-expression of each training sample by taking an information recommendation weight set comprising a constant weight and a plurality of feature weights as a variable according to the value of each model feature of each training sample;

respectively obtaining the actual recommendation score of each training sample to construct a loss function according to the information recommendation sub-expressions and the actual recommendation scores of the training samples;

and solving the loss function, determining the constant weight of the information recommendation weight set and the value of each characteristic weight, and finishing the training of the information recommendation model.

Optionally, the step of separately obtaining the actual recommendation score of each training sample includes:

for each training sample, determining an actual recommendation score of the training sample according to an actual information acquisition behavior of the user corresponding to the training sample on the information corresponding to the training sample;

the actual information acquisition behavior comprises that information acquisition is not carried out, information acquisition is carried out but information is not completely acquired, information is completely acquired and fed back positively and comment information is fed back positively or information is completely acquired and information sources are concerned.

Optionally, the step of constructing a loss function includes:

for each training sample, determining a corresponding loss expression according to the information recommendation sub-expression and the actual recommendation score of the training sample;

and summing the loss expressions of each training sample to obtain the loss function.

Optionally, the step of solving the loss function comprises:

setting the constant weight and the initial value of each characteristic weight included in the information recommendation weight set as random numbers within a preset numerical range;

substituting the information recommendation weight set with the initial value into the loss function to carry out iterative processing;

and when the information recommendation weight set obtained by the iterative processing meets a convergence condition, terminating the iterative processing, and determining the constant weight of the information recommendation weight set and the value of each characteristic weight, otherwise, continuing the iterative processing.

Alternatively,

the convergence condition is that the number of times of the iterative processing is not less than a preset number threshold;

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

the convergence condition is that an iteration result value of the information recommendation weight set obtained by the iteration processing is smaller than a preset result threshold value;

and determining the iteration result value according to a loss function substituted by the information recommendation weight set obtained by the iteration processing, the corresponding constant weight and the result of partial derivation of each characteristic weight.

Optionally, the method further comprises:

and executing the step of training the information recommendation model according to a preset training period.

According to a second aspect of the present invention, there is provided an information recommendation apparatus, comprising:

the candidate information selecting unit is used for selecting a plurality of pieces of candidate information according to historical behavior data related to the user information acquisition behavior;

the information recommendation score acquisition unit is used for respectively acquiring the information recommendation score of each piece of candidate information to the target user according to the acquired information recommendation model;

the information recommendation model is obtained according to training samples related to the user portrait and the information portrait;

and the target information selecting unit is used for selecting the candidate information of which the information recommendation score meets the preset recommendation condition and recommending the candidate information to the target user as target information.

According to a third aspect of the present invention, there is provided an electronic apparatus, comprising:

a memory for storing executable instructions;

and the processor is used for operating the electronic equipment to execute any information recommendation method provided by the first aspect of the invention according to the executable instruction.

According to one embodiment of the disclosure, a plurality of pieces of candidate information are acquired according to historical behavior data related to information acquisition historical behaviors, information recommendation scores of each piece of candidate information for a target user are acquired according to an information portrait of each piece of candidate information and a user portrait of the target user respectively based on an information recommendation model, the candidate information of which the information recommendation scores meet preset recommendation conditions is selected and recommended to the target user as the target information, the acquired state of the candidate information and the information acquisition behaviors of the target user are integrated, the preference degree of the target user for the candidate information is accurately and effectively acquired, the target information which better meets the preference of the user is selected and accurately recommended to the target user, the information acquisition requirements of the target user are actually met, and the information acquisition experience of the target user is improved.

Other features of the present invention and advantages thereof will become apparent from the following detailed description of exemplary embodiments thereof, which proceeds with reference to the accompanying drawings.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description, serve to explain the principles of the invention.

Fig. 1 is a block diagram showing an example of a hardware configuration of an electronic apparatus 1000 that can be used to implement an embodiment of the present invention.

Fig. 2 shows a flowchart of an information recommendation method of an embodiment of the present invention.

FIG. 3 shows a flowchart of the steps of training an information recommendation model of an embodiment of the present invention.

Fig. 4 shows a block diagram of the information recommendation apparatus 3000 of the embodiment of the present invention.

Fig. 5 shows a block diagram of an electronic device 4000 of an embodiment of the invention.

Detailed Description

Various exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that: the relative arrangement of the components and steps, the numerical expressions and numerical values set forth in these embodiments do not limit the scope of the present invention unless specifically stated otherwise.

The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the invention, its application, or uses.

Techniques, methods, and apparatus known to those of ordinary skill in the relevant art may not be discussed in detail but are intended to be part of the specification where appropriate.

In all examples shown and discussed herein, any particular value should be construed as merely illustrative, and not limiting. Thus, other examples of the exemplary embodiments may have different values.

It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, further discussion thereof is not required in subsequent figures.

< hardware configuration >

Fig. 1 is a block diagram showing a hardware configuration of an electronic apparatus 1000 that can implement an embodiment of the present invention.

The electronic device 1000 may be a laptop, desktop, cell phone, tablet, etc. As shown in fig. 1, the electronic device 1000 may include a processor 1100, a memory 1200, an interface device 1300, a communication device 1400, a display device 1500, an input device 1600, a speaker 1700, a microphone 1800, and the like. The processor 1100 may be a central processing unit CPU, a microprocessor MCU, or the like. The memory 1200 includes, for example, a ROM (read only memory), a RAM (random access memory), a nonvolatile memory such as a hard disk, and the like. The interface device 1300 includes, for example, a USB interface, a headphone interface, and the like. The communication device 1400 is capable of wired or wireless communication, for example, and may specifically include Wifi communication, bluetooth communication, 2G/3G/4G/5G communication, and the like. The display device 1500 is, for example, a liquid crystal display panel, a touch panel, or the like. The input device 1600 may include, for example, a touch screen, a keyboard, a somatosensory input, and the like. A user can input/output voice information through the speaker 1700 and the microphone 1800.

The electronic device shown in fig. 1 is merely illustrative and is in no way meant to limit the invention, its application, or uses. In an embodiment of the present invention, the memory 1200 of the electronic device 1000 is configured to store instructions for controlling the processor 1100 to operate to execute any information recommendation method provided by the embodiment of the present invention. It will be appreciated by those skilled in the art that although a plurality of means are shown for the electronic device 1000 in fig. 1, the present invention may relate to only some of the means therein, e.g. the electronic device 1000 relates to only the processor 1100 and the storage means 1200. The skilled person can design the instructions according to the disclosed solution. How the instructions control the operation of the processor is well known in the art and will not be described in detail herein.

< example >

In the present embodiment, an information recommendation method is provided. The information may be any content available to the user, for example, the information may be short videos, pictures, news, self-media articles, and so forth. In one example, the information may be fragmented content provided to the user in the form of a feed stream that is continuously updated and presented to the user content stream, and the information provided in the form of a feed stream may be a picture stream, a short video stream, a news stream, or the like.

As shown in fig. 2, the information recommendation method includes: steps S2100-S2300.

In step S2100, a plurality of pieces of candidate information are acquired based on the historical behavior data related to the information acquisition historical behavior.

The historical behavior data is data related to information acquisition historical behaviors, and the information acquisition historical behaviors may include historical behaviors of acquiring information of any user.

In one example, the historical behavior data includes at least one of a historical information record, an information source record, and an overall information acquisition record.

The history information record comprises information identification of history information which is completely acquired by the target user. The information identifier is used for uniquely identifying a piece of information and can be information code, information ID and the like. In this example, whether the target user completely acquires the information may be determined according to the type of the information, for example, when the information is a picture, the picture is completely acquired when the target user completely opens the picture after clicking to watch the information, when the information is an article, the article is completely acquired when the target user completely opens the article after clicking to read the end of the article, when the information is a video, the video is completely acquired when the target user completely plays the information to the end of the video after clicking, and the like.

The information source record includes a source identification of the information source that the target user has focused on. In this example, the information source is a content source providing information distribution service, and may be a consultation number, a self-media number, or a consultation website. The source identifier is used to uniquely identify the information source, and may be a number, an ID, a name of a holder of the information source, and the like.

The whole information acquisition record comprises the time when each piece of information is acquired by the user in the information database. The information database includes all information that can be acquired by the user, and may be constructed by means of network capture, manual upload, and the like, which is not limited in this example. In this example, each piece of information may be considered to be acquired by the user once it is clicked on by the user or opened by a network jump.

In this example, the step of obtaining the plurality of pieces of candidate information may include: steps S2110, S2120, and S2130.

Step S2110, when the historical behavior data comprises the historical information record, according to the historical information record, obtaining the information preference score of the target user for each piece of information in the information database, and selecting the information of which the descending order of the information preference score meets a preset first order range as candidate information.

The information preference score of the target user for the information embodies the preference degree of the target user for the information, and the information with the descending sorting order of the information preference score conforming to the preset first sorting range is selected as the candidate information, so that the candidate information is the information with higher preference degree of the target user, the target information recommended to the user is selected from the candidate information in the subsequent steps, the preference of the user for the information can be met more accurately and effectively, and the requirement of the user for obtaining the information is met actually.

The preset first sorting range may be set according to a specific application scenario or application requirements, for example, the preset first sorting range may be set to 1-100.

In step S2110, the step of acquiring the information preference score of the target user for each piece of information included in the information database may include:

and summing the information database comprising each piece of information with the information correlation coefficient of each piece of historical information recorded in the historical information record to obtain the information preference score of the information.

The information correlation coefficient is determined according to a first user set related to the information and a second user set corresponding to the historical information, wherein the first user set comprises all users which have completely acquired the information, and the second user set comprises all users which have completely acquired the historical information.

The information database includes all information that can be acquired by the user, and may be constructed by means of network capture, manual upload, and the like, which is not limited in this example.

Assuming that, for a target user U, history information recorded in the history information record that all target users have completely acquired belongs to a set Iu, and for a piece of history information i included in Iu, a second set of users is U_iFor a piece of information j included in the information database, the first set of users is U_j；

For information j, information correlation coefficient Sim with history information i_j,iComprises the following steps:

wherein, | U_j∩U_iIs the first userSet is U_jIs U with the second set of users_iThe number of users in the intersection of (1); i U_j∪U_iI is that the first set of users is U_jIs U with the second set of users_iThe number of users in the union;

for target user u, the information preference score for information j is Like_u,jComprises the following steps:

by analogy, the information preference score of the target user u for each piece of information in the information database can be obtained.

Step S2120, when the historical behavior data includes the information source record, selecting information, which is issued by each information source in the information source record in a preset issuing time period and has been acquired completely, and whose descending order of times accords with a preset second ordering range, as candidate information.

The preset issuing time period and the preset second sorting range may be set according to a specific application scenario or an application requirement, for example, the preset issuing time period is set to the latest week, and the preset second sorting range is set to 1 to 100.

The information source records are recorded with information sources concerned by target users, and the information which is issued within a preset issuing time period and has higher complete acquisition times by the users is selected as candidate information from the information source records concerned by each target user, so that the candidate information is the information which is possibly provided with higher attention by the target users, the target information which is selected from the candidate information in the subsequent steps and is recommended to the users can be combined, the preference of the users for the information can be met more accurately and effectively, and the requirement of the users for acquiring the information is met actually.

Step S2130, when the historical behavior data includes the whole information acquisition record, acquiring, according to the whole information acquisition record, the total times of each piece of information included in the information database, which is acquired by the user in a preset recording time period, to select information whose descending sort order of the total times acquired by the user matches a preset third sort range, as the candidate information.

The information of the preset recording time period and the preset third sorting range may be set according to a specific application scenario or an application requirement, for example, the preset recording time period may be set to the last day, and the preset third sorting range may be set to 1 to 2000.

The whole information acquisition record comprises the time when each piece of information is acquired by the user in the information database, and the total times of acquiring each piece of information in the information database by the user in a preset recording time period can be counted according to the whole information acquisition record. For example, the preset recording period is the last day, and the total number of times acquired by the user may be the total number of clicks of the information on the last day.

The information of which the descending sorting order of the total times acquired by the user meets the preset third sorting range is selected as the candidate information, so that the candidate information is information with high timeliness and hot spot, the target information recommended to the user is selected from the candidate information in the subsequent step, the preference of the user on the information can be met more accurately and effectively, and the requirement of the user for acquiring the information is met actually.

In practical applications, some information with poor content quality (for example, articles with colloquial words and the like) may exist in the information with the descending sort order of the total times acquired by the user and conforming to the preset third sort range, and in this example, the information with poor content quality may not be used as candidate information after being screened manually or by a machine, so as to improve the content quality of the candidate information. In addition, after the information with poor content quality is screened from the information with the total times of descending sort order conforming to the preset third sort range, the information with the preset number (for example, 100) can be randomly selected as the candidate information, so that the processing efficiency is improved.

It should be understood that, in the present embodiment, according to a specific application scenario, the step S2100 may include only one of the steps S2110 to S2130, or may include a plurality of steps S2110 to S2130.

When the step S2100 includes a plurality of steps S2110 to S2130 and there are a plurality of candidate information selection sources, there may be repeated candidate information, and the repeated candidate information may be removed through deduplication processing correspondingly, so as to improve the processing efficiency.

After acquiring a plurality of pieces of candidate information in step S2100, the process proceeds to:

step S2200 is that based on the information recommendation model, the information recommendation score of each candidate information to the target user is obtained according to the information portrait of each candidate information and the user portrait of the target user.

The information recommendation model is a model trained according to a plurality of collected training samples including user portraits and information portraits and is used for obtaining information recommendation scores of information to users.

The information image is image data representing the attribute of the acquired state of the corresponding information. The acquired state of information may include states that information is not acquired, information is not fully acquired, information is fully acquired, and so on.

The user image is image data representing attributes related to information acquisition behavior of a corresponding user. The information acquisition behaviors may include behaviors of not performing information acquisition, performing information acquisition but not completely acquiring information, and giving information feedback, and the information feedback may include positive feedback such as approval, selection of favorite option, negative feedback such as disfavor, poor comment click, and giving text comment information.

Through the information recommendation model, the information recommendation score of the candidate information to the target user is obtained according to the information portrayal of the candidate information and the user portrayal of the target user, the obtained state of the candidate information and the information obtaining behavior of the target user can be integrated, and the preference degree of the target user to the candidate information is accurately and effectively obtained, so that accurate information recommendation is performed on the target user in combination with subsequent steps, the information obtaining requirement of the target user is actually met, and the information obtaining experience of the target user is improved.

In one example, the information recommendation model includes a plurality of model features and an information recommendation weight set. Each model feature has a unique feature identifier, and the feature identifier is used for uniquely identifying the model feature, for example, the feature identifier may be a preset number, or a text feature code may be extracted from each model feature by using, for example, a one hot code (one hot code) to obtain the feature identifier. And each characteristic weight is uniquely corresponding to one model characteristic.

In this example, step S2200 may include: steps S2210-S2220.

Step S2210, determining a value corresponding to each model feature included in the information recommendation model according to the information portrait of the candidate information and the user portrait of the target user.

In this example, the information representation is data acquired within a preset information statistic period, relating to a state in which the corresponding information is acquired by the user. The user representation is representation data acquired within a preset user statistic period and related to the corresponding user information acquisition behavior.

The preset information counting time period and the preset user counting time period can be set according to specific application scenes or application requirements. For example, the preset information statistic period may be set to the last 3 days, and the preset user statistic period may be set to the last 1 month.

And determining a value corresponding to each model feature in the information recommendation model according to the information portrait of the candidate information and the user portrait of the target user, and evaluating by using the information recommendation model and integrating the information acquisition state of the candidate information and the information acquisition behavior of the target user so as to accurately and effectively acquire the preference degree of the target user on the candidate information in combination with subsequent steps.

In a more specific example, the information representation includes at least an average number of full acquisitions, a forward feedback ratio, a full acquisition ratio, and an information label of the corresponding information.

The average number of complete acquisitions is the number of times the information is completely acquired in each preset time unit averaged over the information statistics period. The preset time unit may be set according to a specific application scenario or application requirements, for example, set to one day. How to determine that the information is completely acquired has been described in detail above and is not described in detail here. In this example, the number of times that the acquired information is completely acquired may be counted in the information counting time period, and then divided by the number of preset time units included in the information counting time period, so as to calculate the average complete acquisition time.

The forward feedback rate is a ratio of the number of times of obtaining forward feedback after information is acquired in the information statistics period to the number of times of issuing the information. Information is usually distributed and exposed on an information display platform in the forms of pushing titles to each user, providing access links and the like, and the information can be considered to be distributed once when being exposed to each user. After the information is acquired by a user, if positive feedback such as approval, favorite item selection, higher score giving and the like of the user is acquired, the user can be considered as one positive feedback. In this example, the number of times of obtaining the forward feedback and the number of times of issuing after the obtained information is obtained may be counted within the information counting period, and the forward feedback rate may be calculated.

The complete acquisition ratio is a ratio of the number of times the information is completely acquired within the information statistic period and the number of times the information is issued. How to determine that the information is completely acquired has been described in detail above and is not described in detail here. In this example, the number of times that the acquired information is completely acquired and the number of times of issuance may be counted in the information counting period, and the complete acquisition ratio is calculated.

The information tag is used for identifying the type or the characteristics of the information, can be configured by an information source in a self-defining way when providing the information, and can also be obtained by extracting the characteristics according to the information content.

The user representation includes at least a complete acquisition information record, a positive feedback information record, and an information preference tag of the corresponding user.

The complete acquired information record comprises information identification of information which is acquired by a corresponding user completely and accords with a first preset number in a user counting time period. The first preset number may be set according to a specific application scenario or application requirements, for example, the first preset number is set to 30. The information identifier is used to uniquely identify the information. The information identifying the related content and how to determine that the user completely acquires the information are described above, and are not described herein again.

The forward feedback information record comprises information identifications of information conforming to a second preset number, which are given forward feedback by the corresponding user in the user counting time period. The second preset number may be set according to a specific application scenario or application requirement, for example, the second preset number is set to 30. The information identifier is used to uniquely identify the information. The information identifying the related content and how to determine that the user gives the information forward feedback have been described above and will not be described herein.

The information preference tag is an information tag of which the repetition times meet a preset repetition threshold value in information tags of all information recorded in the complete acquisition information record and/or the forward feedback information record. The repetition threshold may be set according to a specific application scenario or application requirements, for example, the repetition threshold is set to 5.

A plurality of model features included in the information recommendation model respectively belong to a plurality of different feature classes. Each feature class includes at least one model feature. The information recommendation model includes a plurality of different feature classes. The plurality of feature classes include a first feature class cross-correlated with a full acquisition information record of the user and with an information identity of the information, a second feature class cross-correlated with a forward feedback information record of the user and with an information identity of the information, a third feature class cross-correlated with an information preference tag of the user and with an information tag of the information, a fourth feature class correlated with an average number of full acquisitions of the information, a fifth feature class correlated with a forward feedback rate of the information, and a sixth feature class correlated with a full acquisition rate.

The first characteristic class is cross-correlated with the user's complete acquisition information record and with the information identification of the information. For example, the complete acquisition information record of a user includes the information identifier: fid1, fid2, fid 3; one piece of information which can be acquired by the user has an information identifier of id1, and the first feature class includes 3 model features obtained by cross-correlation processing, namely fid1& id1, fid2& id1 and fid3& id 1. By analogy, when the information recommendation model is trained, according to the information portrait and the user portrait included in each obtained training sample, the information identification of the information corresponding to the information portrait and the complete obtained information record included in the user portrait are subjected to the cross correlation processing, and the model features included in the first feature class are constructed.

The second characteristic class is cross-correlated with the user's forward feedback information record and with the information identification of the information. For example, a user's positive feedback information record includes information identifying: l id1, lid2, lid 3; one piece of information which can be acquired by the user has an information identifier of id1, and the second feature class includes 3 model features obtained by cross-correlation processing, namely lid1& id1, lid2& id1 and lid3& id 1. By analogy, when the information recommendation model is trained, according to the information portrait and the user portrait included in each obtained training sample, the information identification of the information corresponding to the information portrait and the forward feedback information record included in the user portrait are subjected to the cross correlation processing, and the model features included in the second feature class are constructed.

The third feature class is cross-correlated with information preference labels of the user and with information labels of the information. For example, a user's information preference label is "lovely pet" or "fun", a user's information label is "dance" or "beauty", and the third feature class includes 4 model features obtained by cross-correlation processing: respectively, "lovely pet & dance", "lovely pet & beauty", "fun & dance", "fun & beauty". By analogy, when the information recommendation model is trained, according to the information portrait and the user portrait included in each obtained training sample, the information labels included in the information portrait and the information preference labels included in the user portrait are subjected to the cross correlation processing, and model features included in the third feature class are constructed.

The fourth feature class is associated with an average number of complete acquisitions of information. The fourth characteristic class can comprise a plurality of model characteristics such as 0-100 average complete acquisition times, 100-500 average complete acquisition times, 500-2000 average complete acquisition times, 2-1 ten thousand average complete acquisition times, 1-10 ten thousand average complete acquisition times, 10-50 ten thousand average complete acquisition times, more than 50 ten thousand average complete acquisition times and the like. Alternatively, when the information recommendation model is trained, the model features included in the fourth feature class may be constructed by performing segmentation division according to a numerical range related to the average complete acquisition frequency included in the information representation included in each acquired training sample.

The fifth feature class is associated with a forward feedback ratio of the information. The fourth feature class can include a plurality of model features such as a forward feedback ratio of 0% -0.5%, a forward feedback ratio of 0.5% -1%, a forward feedback ratio of 1% -3%, a forward feedback ratio of 3% -6%, a forward feedback ratio of 6% -10%, and a forward feedback ratio of more than 10%. Alternatively, when training the information recommendation model, the model features included in the fifth feature class may be constructed by performing segmentation division based on a numerical range related to a forward feedback ratio included in the information representation included in each acquired training sample.

The sixth feature class is associated with a complete acquisition rate of information. The fourth feature class may include a plurality of model features including a complete acquisition ratio of 0% -0.5%, a complete acquisition ratio of 0.5% -1%, a complete acquisition ratio of 1% -3%, a complete acquisition ratio of 3% -6%, a complete acquisition ratio of 6% -10%, and a complete acquisition ratio of 10% or more. Alternatively, when training the information recommendation model, the model features included in the sixth feature class may be constructed by performing segmentation division based on a numerical range related to the complete acquisition ratio included in the information representation included in each acquired training sample.

In this more specific example, the step S2210 of determining, according to the information representation of the candidate information and the user representation of the target user, a value corresponding to each model feature included in the information recommendation model may include: steps S2211-S2212.

And step S2211, setting the value of the model feature as 1 when determining that the model feature included in the information recommendation model is included according to the information portrait of the candidate information and the user portrait of the target user.

The information identification of the candidate information is id1, the average complete acquisition times included in the information portrait is 60, the forward feedback ratio is 0.3, the complete acquisition ratio is 0.6, and the information label is "dance"; the complete acquired information records in the user portrait of the target user comprise fid1 and fid2, the forward feedback information records comprise fid1, and the information preference label is 'laugh';

correspondingly, the model features that can be determined to have include: "id 1& fid 1", "id 1& fid 2" in the first class of feature classes; "id 1& fid 1" in the second class of feature classes; "dance & fun" in the third feature class; "average complete acquisition times 100-500" in the fourth feature class; the values of the model features are correspondingly set to be 1 by '0% -0.5% of the forward feedback ratio' in the fifth feature class and '0.5% -1% of the complete acquisition ratio' in the sixth feature class.

And step S2212, setting the value of the model feature to be 0 when determining that one model feature included in the information recommendation model does not exist according to the information portrait of the candidate information and the user portrait of the target user.

Continuing with the example based on step S2211, for all model features included in the information recommendation model and except the determined model features, the values of the model features may be set to 0, so as to determine the values of all model features included in the information recommendation model.

After determining values of all model features included in the information recommendation model, entering:

and step S2220, acquiring the information recommendation score of the candidate information to the target user according to the information recommendation weight set and the determined value corresponding to each model feature.

The assumption information recommendation model comprises n model characteristics x₁,x₂,......,x_nThe information recommendation weight set comprises a constant weight b and n characteristic weights a₁,a₂,......,a_nAfter n model features x are determined₁,x₂,......,x_nAfter the value of (3), the information recommendation score Y of the candidate information to the target user can be calculated according to the following formula:

by analogy, the information recommendation score of each piece of candidate information for the target user can be obtained according to the steps S2210-S2220.

And step S2300, selecting candidate information of which the information recommendation score meets preset recommendation conditions as target information to be recommended to a target user.

The preset recommendation condition is a condition for judging whether the candidate information is recommended to the user according to the information recommendation score, and can be set according to a specific application requirement or an application scene.

For example, the preset recommendation condition is that the candidate information of which the descending sort order of the information recommendation score meets the preset sort range is the target information.

The preset sequencing range may be set according to specific application requirements or application scenarios, for example, set to 1-1000.

In an example, the information recommendation method provided in this embodiment further includes a step of training and acquiring the information recommendation model, as shown in fig. 3, including: steps S3100-S3500.

Step S3100, according to a plurality of training samples obtained in a statistical period, a plurality of model features included in an information recommendation model are constructed.

The statistical time period may be set according to a specific application scenario or application requirements. For example, the statistical period may be set to the last month.

Each training sample includes a user representation of a user and an information representation of a piece of information.

The user representation is representation data acquired within a preset user statistic period and related to the corresponding user information acquisition behavior. As exemplified above, a user representation may include at least a complete acquisition information record, a positive feedback information record, and an information preference tag for a corresponding user.

The information representation is data acquired within a preset information statistic period and related to a state in which the corresponding information is acquired by a user. As exemplified above, an information representation may include at least an average number of full acquisitions, a forward feedback ratio, a full acquisition ratio, and an information label of corresponding information.

According to the user portrait and the information portrait included in each training sample in the plurality of training samples, a plurality of model features included in the information recommendation model can be constructed and obtained.

For example, a plurality of model features included in the information recommendation model respectively belong to a plurality of different feature classes. Each feature class includes at least one model feature. The information recommendation model includes a plurality of different feature classes. The plurality of feature classes include a first feature class cross-correlated with a full acquisition information record of the user and with an information identity of the information, a second feature class cross-correlated with a forward feedback information record of the user and with an information identity of the information, a third feature class cross-correlated with an information preference tag of the user and with an information tag of the information, a fourth feature class correlated with an average number of full acquisitions of the information, a fifth feature class correlated with a forward feedback rate of the information, and a sixth feature class correlated with a full acquisition rate. The content including the model features in the first, second, third, fourth, fifth and sixth feature classes may be constructed according to the above, and a plurality of feature classes included in the information recommendation model are correspondingly constructed, which is not described herein again.

Step S3200, determining a value of each model feature of each training sample, respectively.

The step of determining the value of each model feature of each training sample may be, for example, determined by the specific implementation in step S2210 according to the information portrait and the user portrait included in the training sample, and the value of each model feature included in the information recommendation model corresponding to the training sample is determined, which is not repeated herein.

And S3300, respectively determining an information recommendation sub-expression of each training sample by taking an information recommendation weight set comprising a constant weight and a plurality of feature weights as a variable according to the value of each model feature of each training sample.

The assumption information recommendation model comprises n model characteristics x₁,x₂,......,x_nDetermining the value of the k training sample to n model features

Then, the information recommendation weight set comprises a constant weight b and n characteristic weights a₁,a₂,......,a_nAs a variable, it can be obtained that the k-th training sample information recommends that the sub-expression is Y_k:

According to the method, the information recommendation sub-expression of each training sample can be determined.

And step S3400, respectively obtaining the actual recommendation score of each training sample, and constructing a loss function according to the information recommendation sub-expression and the actual recommendation score of the plurality of training samples.

In this example, the step of obtaining the actual recommendation score of each training sample respectively may include:

and for each training sample, determining the actual recommendation score of the training sample according to the actual information acquisition behavior of the user corresponding to the training sample on the information corresponding to the training sample.

The actual information acquisition behavior comprises that information acquisition is not carried out, information acquisition is carried out but information is not completely acquired, information is completely acquired and forward feedback is given, information is completely acquired and forward feedback and comment information are given, or information is completely acquired and information sources are concerned.

In this example, different actual recommendation scores may be set according to different actual information acquisition behaviors.

Taking the information corresponding to the training sample as a short video example, in the actual information acquisition behavior: the non-information-acquisition is a non-clicked video, the information acquisition is carried out but the non-complete information acquisition is that the clicked video does not finish playing, the complete information acquisition is that the clicked video finishes playing, the information acquisition is complete and the forward feedback is given that the clicked video finishes playing and the clicked video finishes playing, the information acquisition is complete and the forward feedback is given that the forward feedback and the comment information is that the clicked video finishes playing and commenting, the complete information acquisition and attention information source is that the clicked video finishes playing and the video author pays attention to, and correspondingly, the actual recommendation score can be determined according to the following table:

table 1 set actual recommendation score example

It should be understood that the actual recommendation scores for different actual information obtaining behaviors may be flexibly set according to different types of information, different actual application scenarios, or different application requirements, which is not an example here.

By analogy, based on the above method, the actual recommendation score of each training sample can be obtained.

After obtaining the actual recommendation score of each training sample, the step of constructing the loss function according to the information recommendation sub-expression and the actual recommendation score of the plurality of training samples may include: steps S3410-S3420.

Step S3410, for each training sample, according to the information recommendation sub-expression and the actual recommendation score of the training sample, determining a corresponding loss expression.

Assuming that the number of the collected training samples is m, the obtained actual recommendation score for the k training sample is y_kThe information recommendation sub-expression is Y_kThe corresponding loss expression is (y)_k-Y_k)²(k ═ 1.., m); wherein the content of the first and second substances,

step S3420, sum the loss expressions of each training sample to obtain a loss function.

In this example, the loss function is:

wherein the content of the first and second substances,

and step S3500, solving the loss function, determining the constant weight of the information recommendation weight set and the value of each characteristic weight, and finishing the training of the information recommendation model.

For example, step S3500 may include: steps S3510-S3530.

Step S3510, setting the constant weight included in the information recommendation weight set and the initial value of each feature weight as a random number within a preset numerical range.

Assume information recommendation weight set { b, a₁,a₂,......,a_nComprises a constant weight b and n characteristic weights a₁,a₂,......,a_nThe initial value may be set to a random number of a preset numerical range. The preset value range may be set according to an application scenario or an application requirement, for example, the preset value range is set to 0-1, such that the constant weight b and the n feature weights a₁,a₂,......,a_nAre all random numbers between 0-1.

And S3520, substituting the information recommendation weight set with the initial value into the loss function to perform iterative processing.

For example, step S3520 may include: steps S3521-S3522.

Step S3521, the constant weight and each feature weight are respectively subjected to value obtaining and convergence parameters of the constant weight before the current iteration or the feature weight and a loss function substituted into the information recommendation weight set before the current iteration, and the value of the corresponding constant weight or the feature weight after the iteration is obtained.

The convergence parameter is a relevant parameter for controlling the convergence speed of the iterative process, and may be set according to an application scenario or an application requirement, for example, set to 0.01.

Suppose that the iteration is the (k +1) th iteration (k is the initial value of0, plus 1 with each iteration), the feature weight a is added_nThe value of the feature weight before the iteration is a_n ^(k)The convergence parameter is rho;

substituting the loss function of the information recommendation weight set before the iteration into

Corresponding iterated values a of the feature weights_n ^(k ⁺¹⁾Comprises the following steps:

similarly, for a constant weight b, the value b of the feature weight after the corresponding iteration is^(k+1)Comprises the following steps:

and S3522, obtaining an information recommendation weight set after the iteration according to the constant weight and the value after each characteristic weight iteration.

Assuming that the iteration is the (k +1) th iteration (the initial value of k is 0, and 1 is added along with each iteration), the information recommendation weight set after the iteration is { b, a }₁,a₂,...,a_n}^(k+1)。

And S3530, when the information recommendation weight set obtained by the iterative processing meets the convergence condition, terminating the iterative processing, and determining the constant weight of the information recommendation weight set and the value of each characteristic weight, otherwise, continuing the iterative processing.

In this example, the convergence condition may be set according to a specific application scenario or application requirement.

For example, the convergence condition is that the number of iterative processes is greater than a preset number threshold. The preset time threshold may be set according to engineering experience or experimental simulation results, and may be set to 300, for example. Correspondingly, assuming that the number of iterative processes is k +1, the number threshold is itemnams, and the corresponding convergence condition is: k is not less than itemNums.

For another example, the convergence condition is that an iteration result value of the information recommendation weight set obtained by the iteration processing is smaller than a preset result threshold. The iteration result value is determined by the result of partial derivation of the corresponding constant weight or each characteristic weight according to the loss function substituted by the information recommendation weight set obtained by the iteration processing.

Assume information recommendation weight set { b, a₁,a₂,......,a_nN +1 feature weights and constant weights;

the iteration result value of the (k +1) th iteration is:

the result threshold may recommend the set of weights b, a based on the information₁,a₂,......,a_nsetting, for example, to (n +1) × α, where α may be set according to engineering experience or experimental simulation results, for example, to set α to 0.01 × ρ, and ρ is the convergence parameter described above;

correspondingly, the convergence condition is:

<(n+1)×α

in an example, the convergence condition is that any one of the convergence conditions in the two examples is satisfied, and the specific convergence condition has been described in the two examples and is not described herein again.

Suppose that the information recommendation weight set { b, a) obtained by the (k +1) th iteration processing₁,a₂,...,a_n}^(k+1)When the convergence condition is met, stopping the iterative processing to obtain all the corresponding a_i ^(k+1)(i ═ 1.., n) and b^(k+1)And taking values, otherwise, continuing the iterative processing until the information recommendation weight set meets the convergence condition.

In practical application, the content and the quantity of information that can be acquired by a user may change with time, the state of each piece of information that is acquired by the user may also change with time, and the behavior of the user acquiring information may also change with time, and these factors may affect the actual recommendation effect on the target user recommendation information, so in this embodiment, the step of training the information recommendation model shown in fig. 3 may also be performed on the information recommendation model according to a preset training period, so that the information recommendation model is adaptive to the change of the time according to the factor that adapts to the above-mentioned influence on the information recommendation effect, and the preference of the target user on the information is more accurately acquired, so as to implement information recommendation that actually meets the requirements of the target user. The training period may be set according to a specific application scenario or application requirements, for example, to 1 day.

< information recommendation apparatus >

In the present embodiment, there is provided an information recommendation apparatus 3000, as shown in fig. 4, including: the candidate information obtaining unit 3100, the information recommendation score obtaining unit 3200, and the target information selecting unit 3300 are configured to implement any one of the information recommendation methods provided in this embodiment, and details are not repeated here.

A candidate information selecting unit 3100, configured to select a plurality of pieces of candidate information according to historical behavior data related to the user information obtaining behavior.

In one example, the historical behavior data at least includes one of a historical information record, an information source record and an overall information acquisition record; the historical information record comprises information identification of the historical information which is completely acquired by the target user; the information source record comprises a source identifier of an information source concerned by the target user; the whole information acquisition record comprises the time when each piece of information is acquired by the user in the information database;

the candidate information selecting unit 3100 includes:

means for selecting, when the historical behavior data includes the information source record, information that is issued by each information source recorded in the information source record within a preset issuing period and has been completely acquired for a time in a descending order of the number of times in accordance with a preset second order range, as the candidate information;

and when the historical behavior data comprises an overall information acquisition record, acquiring the total times of each piece of information in the information database acquired by the user in a preset recording time period according to the overall information acquisition record, and selecting the information of which the descending sorting order of the total times acquired by the user accords with a preset third sorting range as the candidate information.

In this example, the means for acquiring, when the historical behavior data includes the historical information record, an information preference score of the target user for each piece of information included in an information database according to the historical information record, and selecting the information whose descending order of the information preference score meets a preset first order range, as the candidate information, may include:

a device for summing information correlation coefficients of each piece of historical information recorded by the information and each piece of information included in the historical information in an information database to obtain an information preference score of the information;

The information recommendation score obtaining unit 3200 is configured to obtain information recommendation scores of each piece of candidate information for the target user according to the obtained information recommendation model;

the information recommendation model is obtained according to training samples related to the user portrait and the information portrait.

In one example, the information recommendation model includes a plurality of model features and an information recommendation weight set, each model feature has a unique feature identifier, the information recommendation weight set includes a constant weight and a plurality of feature weights, and each feature weight uniquely corresponds to one model feature.

The information recommendation score obtaining unit 3200 may further include:

a device for determining a value corresponding to each model feature in the information recommendation model according to the information portrait of the candidate information and the user portrait of the target user;

and the device is used for acquiring the information recommendation score of the candidate information to the target user according to the information recommendation weight set and the determined value corresponding to each model feature.

For example, the information representation includes at least an average complete acquisition number, a forward feedback ratio, a complete acquisition ratio, and an information label of the corresponding information; the average complete acquisition times are times of complete acquisition of the information in each preset time unit in the information statistics period; the forward feedback rate is the ratio of the number of times of obtaining forward feedback after the information is acquired in the information counting time period to the number of times of issuing the information; the complete acquisition ratio is a ratio of the number of times the information is completely acquired within the information statistics period and the number of times the information is issued;

the device for determining the value corresponding to each model feature in the information recommendation model according to the information portrait of the candidate information and the user portrait of the target user comprises:

means for setting a value of the model feature to 1 when determining that there is a model feature included in the information recommendation model based on the information representation of the candidate information and the user representation of the target user;

And a target information selecting unit 3300, configured to select the candidate information whose information recommendation score meets a preset recommendation condition, and recommend the candidate information as target information to the target user.

In one example, the preset recommendation condition is that the candidate information of which the descending sorting order of the information recommendation score meets a preset sorting range is the target information.

The information recommendation device 3000 may further include a device for training and acquiring the information recommendation model, including:

the device is used for constructing a plurality of model characteristics included in the information recommendation model according to a plurality of training samples acquired in a statistical time period;

means for determining a value of each of the model features for each training sample, respectively;

a device for determining an information recommendation sub-expression of each training sample respectively by using an information recommendation weight set comprising a constant weight and a plurality of feature weights as a variable according to the value of each model feature of each training sample;

the device is used for respectively obtaining the actual recommendation score of each training sample so as to construct a loss function according to the information recommendation sub-expression and the actual recommendation score of the training samples;

and the device is used for solving the loss function, determining the constant weight of the information recommendation weight set and the value of each characteristic weight, and finishing the training of the information recommendation model.

In an example, the apparatus for determining the information recommendation partial expression of each training sample respectively by using an information recommendation weight set including a constant weight and a plurality of feature weights as variables according to the value of each model feature of each training sample includes:

means for determining, for each of the training samples, an actual recommendation score for the training sample based on actual information acquisition behavior of the user corresponding to the training sample with respect to the information corresponding to the training sample;

means for determining a corresponding loss expression for each training sample according to the information recommendation sub-expression and the actual recommendation score of the training sample;

and means for summing the loss expressions of each of the training samples to obtain the loss function.

In one example, the apparatus for solving the loss function, determining the constant weight of the information recommendation weight set and the value of each feature weight, and completing the training of the information recommendation model at this time includes:

means for setting the constant weight and the initial value of each feature weight included in the information recommendation weight set to a random number within a preset numerical range;

the device is used for substituting the information recommendation weight set with the initial value into the loss function to carry out iterative processing;

and the device is used for terminating the iterative processing when the information recommendation weight set obtained by the iterative processing meets a convergence condition, determining the constant weight of the information recommendation weight set and the value of each characteristic weight, and if not, continuing the iterative processing.

Further, the convergence condition is that the number of times of the iterative processing is not less than a preset number threshold;

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

In one example, the means for training the information recommendation model further comprises:

means for performing the step of training the information recommendation model according to a preset training period.

It will be appreciated by those skilled in the art that the information recommendation device 3000 can be implemented in various ways. For example, the information recommendation apparatus 3000 may be implemented by an instruction configuration processor. For example, the information recommendation apparatus 3000 may be implemented by storing instructions in a ROM and reading the instructions from the ROM into a programmable device when starting up the device. For example, the information recommendation device 3000 may be solidified into a dedicated device (e.g., ASIC). The information recommendation apparatus 3000 may be divided into units independent of each other, or may be implemented by combining them together. The information recommendation device 3000 may be implemented by one of the various implementations described above, or may be implemented by a combination of two or more of the various implementations described above.

In this embodiment, the information recommendation apparatus 3000 may provide any software product or application program of the information recommendation service, for example, a feed stream software product or application program that provides information in a feed stream format.

< electronic apparatus >

In this embodiment, an electronic device 4000 is further provided, as shown in fig. 5, including:

a memory 4100 for storing executable instructions;

a processor 4200, configured to execute the electronic device to perform any one of the information recommendation methods provided in this embodiment according to the control of the executable instructions.

In this embodiment, the electronic device 4000 may be a mobile phone, a tablet computer, a palm computer, a desktop computer, a notebook computer, or the like. In one example, the electronic device 4000 may be a mobile phone installed with an application (e.g., a feed stream-based application) that provides an information acquisition service. Alternatively, the electronic device 4000 may also be a server in which an application providing the information acquisition service is located. Alternatively, the electronic device 4000 may further include a mobile phone that installs an application providing the information acquisition service and a server in which the application is located.

In this embodiment, the electronic device 4000 may further include other hardware modules, for example, the electronic device 1000 shown in fig. 1.

The information recommendation method, device and electronic equipment provided in the embodiment have been described above with reference to the drawings and examples, by acquiring a plurality of pieces of candidate information according to historical behavior data related to information acquisition historical behaviors, acquiring information recommendation scores of each piece of candidate information to a target user according to an information portrait of each piece of candidate information and a user portrait of the target user respectively based on an information recommendation model, selecting the candidate information of which the information recommendation scores meet preset recommendation conditions to be recommended to the target user as the target information, realizing the comprehensive acquisition state of the candidate information and the information acquisition behaviors of the target user, accurately and effectively acquiring the preference degree of the target user to the candidate information, and target information which better accords with user preferences is selected to carry out accurate recommendation on the target user, the information acquisition requirements of the target user are actually met, and the information acquisition experience of the target user is improved.

The present invention may be a system, method and/or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions embodied therewith for causing a processor to implement various aspects of the present invention.

The computer readable storage medium may be a tangible device that can hold and store the instructions for use by the instruction execution device. The computer readable storage medium may be, for example, but not limited to, an electronic memory device, a magnetic memory device, an optical memory device, an electromagnetic memory device, a semiconductor memory device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a Static Random Access Memory (SRAM), a portable compact disc read-only memory (CD-ROM), a Digital Versatile Disc (DVD), a memory stick, a floppy disk, a mechanical coding device, such as punch cards or in-groove projection structures having instructions stored thereon, and any suitable combination of the foregoing. Computer-readable storage media as used herein is not to be construed as transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission medium (e.g., optical pulses through a fiber optic cable), or electrical signals transmitted through electrical wires.

The computer-readable program instructions described herein may be downloaded from a computer-readable storage medium to a respective computing/processing device, or to an external computer or external storage device via a network, such as the internet, a local area network, a wide area network, and/or a wireless network. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers and/or edge servers. The network adapter card or network interface in each computing/processing device receives computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in a computer-readable storage medium in the respective computing/processing device.

The computer program instructions for carrying out operations of the present invention may be assembler instructions, Instruction Set Architecture (ISA) instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The computer-readable program instructions may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider). In some embodiments, aspects of the present invention are implemented by personalizing an electronic circuit, such as a programmable logic circuit, a Field Programmable Gate Array (FPGA), or a Programmable Logic Array (PLA), with state information of computer-readable program instructions, which can execute the computer-readable program instructions.

Aspects of the present invention are described herein with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and/or block diagrams, and combinations of blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer-readable program instructions.

These computer-readable program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions/acts specified in the flowchart and/or block diagram block or blocks. These computer-readable program instructions may also be stored in a computer-readable storage medium that can direct a computer, programmable data processing apparatus, and/or other devices to function in a particular manner, such that the computer-readable medium storing the instructions comprises an article of manufacture including instructions which implement the function/act specified in the flowchart and/or block diagram block or blocks.

The computer readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other devices to produce a computer implemented process such that the instructions which execute on the computer, other programmable apparatus or other devices implement the functions/acts specified in the flowchart and/or block diagram block or blocks.

The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which comprises one or more executable instructions for implementing the specified logical function(s). In some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions. It is well known to those skilled in the art that implementation by hardware, by software, and by a combination of software and hardware are equivalent.

Having described embodiments of the present invention, the foregoing description is intended to be exemplary, not exhaustive, and not limited to the embodiments disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen in order to best explain the principles of the embodiments, the practical application, or improvements made to the technology in the marketplace, or to enable others of ordinary skill in the art to understand the embodiments disclosed herein. The scope of the invention is defined by the appended claims.

Claims

1. An information recommendation method, comprising:

2. The method of claim 1, wherein,

the historical behavior data at least comprises one of historical information records, information source records and whole information acquisition records; the historical information record comprises information identification of the historical information which is completely acquired by the target user; the information source record comprises a source identifier of an information source concerned by the target user; the whole information acquisition record comprises the time when each piece of information is acquired by the user in the information database;

the step of acquiring a plurality of pieces of candidate information includes:

3. The method of claim 2, wherein,

when the historical behavior data includes the historical information record, the step of obtaining the information preference score of the target user for each piece of information included in the information database according to the historical information record includes:

4. The method of claim 1, wherein,

the information recommendation model comprises a plurality of model features and an information recommendation weight set, each model feature has a unique feature identifier, the information recommendation weight set comprises a constant weight and a plurality of feature weights, and each feature weight is uniquely corresponding to one model feature;

5. The method of claim 4, wherein,

6. The method of claim 1, wherein,

the preset recommendation condition is that the candidate information of which the descending sorting order of the information recommendation score meets a preset sorting range is the target information.

7. The method of claim 1, wherein the method further comprises the step of training the acquisition of the information recommendation model, comprising:

8. The method of claim 7, wherein the step of separately obtaining the actual recommendation score for each of the training samples comprises:

9. The method of claim 7, wherein the step of constructing a loss function comprises:

10. The method of claim 9, wherein the step of solving the loss function comprises:

11. The method of claim 10, wherein,

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

12. The method of claim 7, further comprising:

13. An information recommendation apparatus, comprising:

14. An electronic device, comprising:

a memory for storing executable instructions;

a processor configured to execute the electronic device to perform the information recommendation method according to any one of claims 1 to 12 according to the executable instruction.