CN111382346B - Method and system for recommending content - Google Patents
Method and system for recommending content Download PDFInfo
- Publication number
- CN111382346B CN111382346B CN201811621052.5A CN201811621052A CN111382346B CN 111382346 B CN111382346 B CN 111382346B CN 201811621052 A CN201811621052 A CN 201811621052A CN 111382346 B CN111382346 B CN 111382346B
- Authority
- CN
- China
- Prior art keywords
- candidate content
- content
- user
- candidate
- recommended
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
A method and system for recommending content are provided, the method comprising: generating a prediction sample corresponding to candidate content aiming at the candidate content; predicting the probability of clicking the candidate content after being recommended to the user facing the candidate content according to the prediction sample, wherein the probability is used as the estimated click rate of the candidate content; acquiring optimization index information for measuring the recommended degree of the candidate content; and determining a recommendation index that recommends the candidate content to the user based on the estimated click rate along with the optimization index information. By adopting the method and the system for recommending the content according to the exemplary embodiment of the invention, the recommendation can be more accurately performed for the user, and the income of the content provider can be improved.
Description
Technical Field
The present invention relates generally to content recommendation, and more particularly, to a method and system for recommending content.
Background
With the development of internet technology, massive content is generated, and the recommendation of the interested personalized content to the user can help to greatly improve the use experience of the user and bring great benefit to the content provider. In general, when recommending a specific content to a user, a content operator may desire to predict a possibility that the specific content is accepted by the user after the recommendation, and then perform content recommendation accordingly according to the prediction result. However, the last accepted behavior of the user is itself difficult to collect in some cases, and the click rate as to whether the user accepts may lead to many inaccurate predictions.
Disclosure of Invention
It is an object of exemplary embodiments of the present invention to provide a method and system for recommending content, which overcome at least one of the above-mentioned disadvantages.
According to an aspect of exemplary embodiments of the present invention, there is provided a method for recommending content, including: generating a prediction sample corresponding to candidate content aiming at the candidate content; predicting the probability of clicking the candidate content after being recommended to the user facing the candidate content according to the prediction sample, wherein the probability is used as the estimated click rate of the candidate content; acquiring optimization index information for measuring the recommended degree of the candidate content; and determining a recommendation index that recommends the candidate content to the user based on the estimated click rate along with the optimization index information.
Further, the step of generating the prediction samples corresponding to the candidate content may include: acquiring at least one attribute information corresponding to the candidate content, wherein the at least one attribute information can comprise attribute information related to the candidate content, a user and/or an environment to which the candidate content is directed; a prediction sample corresponding to the candidate content is generated based at least on the at least one acquired attribute information.
Further, the optimization index information may include historical statistics representing actions of the candidate content that are further accepted after being clicked and/or price information measuring benefits obtained after the candidate content is recommended to the user.
Further, the historical statistics may include a retention rate of the candidate content counted during a predetermined period of time after being clicked, wherein the retention rate may refer to a probability that each user reserves a contact address for the candidate content after clicking the candidate content.
Further, the retention rate may be obtained by: obtaining a retention rate of the candidate content from a provider of the candidate content; and/or predicting the retention rate of the candidate content after being clicked for the candidate content.
Further, the method may further comprise: normalizing price information measuring a return obtained after the candidate content is recommended to a user, wherein determining a recommendation index for recommending the candidate content to the user based on the estimated click rate together with the optimization index information may include: a recommendation index for recommending the candidate content to the user is determined based on the estimated click rate and at least one of the historical statistics and normalized price information.
Further, the recommendation index may correspond to a product of the estimated click rate and at least one of the historical statistics and normalized price information.
Further, predicting, as the estimated click rate of the candidate content, a probability that the candidate content is clicked after being recommended to the user according to the prediction sample may include: and predicting the probability of the candidate content being clicked after being recommended to the user according to the prediction sample by using a machine learning model, wherein the probability is used as the estimated click rate of the candidate content.
Further, the machine learning model may be trained by: acquiring at least one attribute information corresponding to a real recommended content which is historically recommended; generating features of the training samples corresponding to the real recommended content based on at least the acquired at least one attribute information in the same feature processing manner as the features of the prediction samples; taking the result of whether the real recommended content is clicked or not as a mark of a training sample; and training the machine learning model based on a set of a plurality of training samples.
Further, the method may further comprise: recommending the candidate content to the user based on the recommendation index by one of: recommending the candidate content with the highest recommendation index in the candidate content collection to a user; a list of recommended content is created, which is provided to a user, wherein the recommended content in the list may be a top-ranked candidate content of a set of candidate content.
Further, the set of candidate content may include all candidate content or at least a portion of candidate content recalled for the characteristics of the faced user.
According to an aspect of exemplary embodiments of the present invention, there is provided a system for recommending content, including: a prediction sample generation unit that generates a prediction sample corresponding to candidate content with respect to the candidate content; the estimated click rate determining unit predicts the probability of the candidate content being clicked after being recommended to the user facing the candidate content according to the prediction sample, and takes the probability as the estimated click rate of the candidate content; an optimization index information acquisition unit for acquiring optimization index information for measuring the recommended degree of the candidate content; and a recommendation index determining unit that determines a recommendation index that recommends the candidate content to the user based on the estimated click rate together with the optimization index information.
Further, the prediction sample generation unit may acquire at least one attribute information corresponding to the candidate content, and generate a prediction sample corresponding to the candidate content based at least on the acquired at least one attribute information, wherein the at least one attribute information may include attribute information related to the candidate content, a user and/or an environment to which the candidate content is directed.
Further, the optimization index information may include historical statistics representing actions of the candidate content that are further accepted after being clicked and/or price information measuring benefits obtained after the candidate content is recommended to the user.
Further, the historical statistics may include a retention rate of the candidate content counted during a predetermined period of time after being clicked, wherein the retention rate may refer to a probability that each user reserves a contact address for the candidate content after clicking the candidate content.
Further, the optimization index information acquisition unit may acquire the electricity retention rate by: obtaining a retention rate of the candidate content from a provider of the candidate content; and/or predicting the retention rate of the candidate content after being clicked for the candidate content.
Further, the recommendation index determining unit may further normalize price information measuring a return obtained after the candidate content is recommended to the user, and determine a recommendation index recommending the candidate content to the user based on the estimated click rate and at least one of the historical statistical information and the normalized price information.
Further, the recommendation index may correspond to a product of the estimated click rate and at least one of the historical statistics and normalized price information.
Further, the estimated click rate determination unit may predict, as the estimated click rate of the candidate content, a probability that the candidate content is clicked after being recommended to the user with respect to the prediction sample using a machine learning model.
Further, the system may further comprise a model training unit, the machine learning model may be trained by: acquiring at least one attribute information corresponding to a real recommended content which is historically recommended; generating features of the training samples corresponding to the real recommended content based on at least the acquired at least one attribute information in the same feature processing manner as the features of the prediction samples; taking the result of whether the real recommended content is clicked or not as a mark of a training sample; and training the machine learning model based on a set of a plurality of training samples.
Further, the system may further include: a content recommendation unit that recommends the candidate content to the user based on the recommendation index by one of: recommending the candidate content with the highest recommendation index in the candidate content collection to a user; a list of recommended content is created, which is provided to a user, wherein the recommended content in the list may be a top-ranked candidate content of a set of candidate content.
Further, the set of candidate content may include all candidate content or at least a portion of candidate content recalled for the characteristics of the faced user.
According to another aspect of exemplary embodiments of the present invention, there is provided a system comprising at least one computing device and at least one storage device storing instructions, wherein the instructions, when executed by the at least one computing device, cause the at least one computing device to perform any of the above methods.
According to another aspect of exemplary embodiments of the present invention, a computer-readable storage medium storing instructions is provided, wherein the instructions, when executed by at least one computing device, cause the at least one computing device to perform any of the above methods.
According to the method and the system for recommending the content, which are disclosed by the embodiment of the invention, the recommendation can be more accurately performed for the user, and the benefit of a content provider can be improved.
Additional aspects and/or advantages of the present general inventive concept will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the general inventive concept.
Drawings
The foregoing and other objects, features, and advantages of exemplary embodiments of the invention will become more apparent from the following detailed description, taken in conjunction with the accompanying drawings that illustrate exemplary embodiments in which:
FIG. 1 illustrates a flowchart of a method for recommending content according to an exemplary embodiment of the present invention;
FIG. 2 shows a flowchart of steps for generating prediction samples corresponding to candidate content, according to an exemplary embodiment of the present invention;
FIG. 3 shows a flowchart of steps for training a machine learning model according to an exemplary embodiment of the present invention;
fig. 4 illustrates a block diagram of a system for recommending content according to an exemplary embodiment of the present invention.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like reference numerals refer to the like elements throughout. The embodiments will be described below in order to explain the present invention by referring to the figures.
Fig. 1 illustrates a flowchart of a method for recommending content according to an exemplary embodiment of the present invention. Here, the method may be performed by a computer program, or may be performed by a dedicated system or computing device for recommending content, as examples.
Referring to fig. 1, in step S10, a prediction sample corresponding to candidate content is generated for the candidate content.
Here, the candidate content may refer to content to be recommended by the user, for example, music, news, advertisements, video, information, lessons, and the like. The step of generating prediction samples corresponding to candidate contents is described below with reference to fig. 2.
Fig. 2 shows a flowchart of the steps of generating a prediction sample corresponding to candidate content according to an exemplary embodiment of the present invention.
Referring to fig. 2, at least one attribute information corresponding to candidate contents is acquired in step S101. Here, the at least one attribute information may include attribute information related to the candidate content, the user and/or the environment to which it is directed.
For example, attribute information related to candidate content may include, but is not limited to, at least one of: the ID of the candidate content, the category of the candidate content, the title of the candidate content, the abstract of the candidate content, the keywords of the candidate content, the part of speech of the keywords of the candidate content, the location where the candidate content is presented to the user.
For example, attribute information related to a user for which candidate content is targeted may include, but is not limited to, at least one of: the gender of the user, the age of the user, the ID of the user, the set of recommended content clicked by the user, the set of recommended content publishers clicked by the user.
For example, the attribute information related to the environment may refer to additional feature information related to the candidate content, the environment in which the user and/or other party are located. As an example, the attribute information related to the environment may include, but is not limited to, at least one of: browser version of the presentation candidate, category of terminal device (e.g., desktop, tablet, smart phone) presenting the candidate, model of terminal device, weather, season, recent hot event.
In an example, the candidate content may be a candidate recommended lesson that may be presented in association in an article of a blog (blog), in which case the relevant information of the blog may include, but is not limited to, at least one of: blog ID, blog title, blog content, number of times a blog is clicked, blog publisher information, time of blog publication candidate content, category of blog, tag of blog.
Alternatively, the at least one attribute information of the candidate content may be acquired periodically or periodically, or may be acquired at a specific point in time or under the trigger of a specific event. Here, the attribute information of the candidate contents may be obtained in batches, wherein each batch performs attribute information collection for a plurality of candidate contents.
In step S102, a prediction sample corresponding to the candidate content is generated based at least on the acquired at least one attribute information.
For example, the obtained at least one attribute information may be regarded as at least one attribute field of the sample instance, and accordingly, the feature portion of the predicted sample may be obtained by performing feature processing on the at least one attribute field. As an example, other additional information, such as timing statistics, etc., may also be incorporated in the above feature processing.
The engineering of the features may be performed in a variety of suitable ways, and the invention is not limited in this regard. For example, various processes, such as operations, combinations, discretizations, serialization, etc., and combinations thereof, may be performed on the attribute fields to generate features of the predicted samples.
Returning to fig. 1, in step S20, the probability that the candidate content is clicked after being recommended to the user to which it is directed is predicted as the estimated click rate of the candidate content according to the prediction sample.
Here, exemplary embodiments of the invention may employ any suitable manner to predict click through rates, such as using human rules, data analysis, or machine learning models, etc. As an example, the estimated click rate of the candidate content may be predicted by a method such as tag similarity calculation, statistical method, or the like.
In an alternative embodiment, the prediction samples may be input to a machine learning model (e.g., logistic regression (logistic regression) model, gradient-enhanced decision tree (Gradient Boosting Decision Tree) model, neural network model) by which the probability of candidate content being clicked after being recommended to the respective user is predicted.
Here, the machine learning model for predicting the click rate of the candidate content may be a machine learning model that the method execution subject itself previously trained, or a trained machine learning model may be directly or indirectly acquired from the outside by the subject. The steps of training a machine learning model according to an exemplary embodiment of the present invention are described below with reference to fig. 3.
FIG. 3 shows a flowchart of the steps of training a machine learning model according to an exemplary embodiment of the present invention.
Referring to fig. 3, at least one attribute information corresponding to a real recommended content that is historically recommended is acquired in step S301.
Here, the acquired at least one attribute information corresponding to the actual recommended content is the same as the at least one attribute information corresponding to the candidate content acquired in step S101. In this step, a number of click conditions after the real content has been historically recommended to the individual users may be obtained for use in forming training samples for the machine learning model.
In step S302, the features of the training samples corresponding to the actual recommended content are generated based on at least the acquired at least one attribute information in the same feature processing manner as the features of the prediction samples are generated.
In step S303, the result of whether the real recommended content is clicked or not is taken as a mark of the training sample.
For example, at least one attribute information corresponding to the real recommended content and a result of whether the real recommended content is clicked or not may be collected in a random recommendation manner to generate the feature of the training sample and the mark of the training sample.
In step S304, a machine learning model is trained based on a set of a plurality of training samples.
Here, any supervised algorithm may be employed to train the model for predicting click through rates, as the present invention is not limited in this regard.
Optionally, the method for recommending content according to an exemplary embodiment of the present invention may further include: updating the machine learning model.
For example, the machine learning model may be updated by: the actual results of whether the candidate content was clicked after being recommended to the user to which it was targeted are recorded to form training samples for model updating based on at least one attribute information corresponding to the candidate content and the actual results of whether the candidate content was clicked after being recommended to the user to which it was targeted, which samples can be used to retrain or incrementally train the machine learning model. By continuously updating the machine learning model, the accuracy of content recommendation can be effectively improved, and more accurate personalized recommendation service can be provided for users.
Returning to fig. 1, in step S30, optimization index information for measuring the degree to which the candidate content is worth being recommended is acquired. Here, the optimization index information may reflect whether the corresponding candidate content is worth actually recommended from various aspects such as a true acceptance rate (e.g., information that the user has clicked but not actually browsed, etc.), a profit situation, etc.
Alternatively, the optimization index information may refer to a subsequent index directly related to the candidate content being clicked, for example, the optimization index information may include historical statistics representing actions that the candidate content is further accepted after being clicked and/or price information measuring the benefits obtained after the candidate content is recommended to the user.
In an alternative embodiment, the historical statistics may include the retention rate of candidate content counted over a predetermined period of time after being clicked. Here, the retention rate may refer to a probability that each user historically reserves a contact (e.g., phone number, mailbox, account number, etc.) for candidate content after clicking on the candidate content. After the candidate recommended course is clicked, whether the user reserves the contact way or not will relate to whether the candidate content will be further accepted by the user, that is, the higher the retention rate of the candidate content, the greater the likelihood that the candidate content will be further accepted, and the more worth recommending the candidate content to the user.
As an example, the retention rate of candidate content may be obtained from a provider of the candidate content. For the case where the candidate content is a candidate recommended course, the retention rate counted by each candidate recommended course in a specific period of time may be collected from the provider of the candidate recommended courses. However, the present invention is not limited thereto, and for example, the retention rate of the candidate content after being clicked may be predicted for the candidate content, that is, a prediction operation mechanism for the retention rate may be established, however, in the case where it is difficult to collect corresponding data regarding whether or not the reality of the contact is left, exemplary embodiments of the present invention may obtain the retention rate by means of statistical information or the like. It is also conceivable to combine the two to some extent.
Aiming at the situation that the candidate content is a candidate recommended course, the electricity retention rate and the course price of the candidate recommended course are two key factors related to whether the candidate recommended course is worth recommending, so that when the candidate content is recommended to a user, the electricity retention rate and the course price can be used as optimization index information, and the aim of improving the overall income of the course is achieved.
In step S40, a recommendation index that recommends candidate content to the user is determined based on the estimated click rate of the candidate content together with the optimization index information.
Here, the estimated click rate and the optimization index information can be simultaneously considered to determine the size of the recommendation index, and the degree and the mode of influencing the recommendation index by the estimated click rate and the optimization index information can be set according to requirements.
Optionally, the method for recommending content according to an exemplary embodiment of the present invention may further include: and normalizing price information for measuring the benefits obtained after the candidate content is recommended to the user.
For example, the price information may be z-score normalized, and then normalized between (0-1) using min-max, so that it can be calculated in the same space. However, the present invention is not limited thereto, and the normalization processing may be performed on the price information using the min-max normalization processing or the z-score normalization processing alone. In addition, other normalization processing methods can be adopted in the field to normalize the price information.
In this case, the recommendation index for recommending the candidate content to the user may be determined based on the estimated click rate of the candidate content and at least one of the history statistical information indicating the action of the candidate content being further accepted after being clicked and the price information after normalization processing.
As an example, the recommendation index may correspond to a product of the estimated click rate and at least one of the historical statistics and the normalized price information.
For example, taking the example that the optimization index information includes the electricity retention rate and the price information, the recommendation index may be determined by the following formula:
δ i ＝μ i ×η i ×τ i (1)
in the formula (1), i represents the i-th candidate content, δ i Recommendation index, μ representing recommendation of the i-th candidate content to the user i Representing the estimated click rate (which may be the output result of the model directly or a-th power of the output result, where a may be an integer between 0 and 10, for example), η of the i-th candidate content i Representing the retention rate of the i candidate content counted in a preset time period after being clicked, tau i Price information representing the benefits obtained after the normalized measurement of the ith candidate content is recommended to the user. Here, 1.ltoreq.i.ltoreq.n, where n is the number of candidate contents. Here, as an example, the set of candidate content may include all candidate content or at least a portion of candidate content recalled for the characteristics of the user at hand.
Optionally, the method for recommending content according to an exemplary embodiment of the present invention may further include: candidate content is recommended to the user based on the recommendation index.
In one example, the candidate content with the highest recommendation index among the set of candidate contents may be recommended to the user.
For example, for each candidate content in the set of candidate contents, a recommendation index for recommending each candidate content to the user is determined separately, so that the candidate content with the highest recommendation index is recommended to the user.
In another example, a list of recommended content may be created to provide the user with the list of recommended content created. Here, the recommended content in the list may be the top-ranked candidate content of the recommendation index among the set of candidate contents.
For example, for each candidate content in the set of candidate contents, a recommendation index for recommending each candidate content to the user may be determined separately, and sorted in descending order according to the recommendation index, and a predetermined number of candidate contents with the top sorted recommendation index are selected to be recommended to the user in list form.
Alternatively, the above-described method for recommending content may be performed on a server of a publisher of the candidate content, for example, the above-described method for recommending content may be performed on a server of a web page that publishes the candidate content, or may also be performed on a server of an application that publishes the candidate content. In addition, the method for recommending content may be performed on a third party server to obtain a recommendation index for recommending the candidate content to the user, and then the publisher of the candidate content recommends the candidate content to the user according to the recommendation index.
According to the method for recommending contents according to the exemplary embodiment of the present invention, the recommendation index is determined in combination with optimization index information such as the retention rate and/or price information of the candidate contents on the basis of predicting the estimated click rate of the candidate contents, so that the recommended candidate contents are more worth being recommended, for example, the possibility of being finally accepted/purchased by the user is greater or the yield is higher, which may bring greater profits to the provider of the candidate contents.
Fig. 4 illustrates a block diagram of a system for recommending content according to an exemplary embodiment of the present invention. Here, the method shown in fig. 1 may be performed by the system shown in fig. 4, as an example.
As shown in fig. 4, a system for recommending content according to an exemplary embodiment of the present invention includes: a prediction sample generation unit 10, a predicted click rate determination unit 20, an optimization index information acquisition unit 30, and a recommendation index determination unit 40.
Specifically, the prediction sample generation unit 10 generates, for candidate contents, a prediction sample corresponding to the candidate contents.
The prediction sample generation unit 10 may acquire at least one attribute information corresponding to the candidate content and generate a prediction sample corresponding to the candidate content based at least on the acquired at least one attribute information. Here, the at least one attribute information may include attribute information related to the candidate content, the user and/or the environment to which it is directed.
The estimated click rate determination unit 20 predicts the probability that the candidate content is clicked after being recommended to the user to which it is directed, as the estimated click rate of the candidate content, based on the prediction sample.
In an alternative embodiment, the estimated click rate determining unit 20 may predict, as the estimated click rate of the candidate content, the probability that the candidate content is clicked after being recommended to the user for the prediction sample using a machine learning model.
Here, the machine learning model for predicting the click rate of the candidate content may be a machine learning model trained in advance by the recommender system itself, or a trained machine learning model may be obtained directly or indirectly from the outside by the recommender system.
In case of training the machine learning model, the system for recommending content according to an exemplary embodiment of the present invention may further include a model training unit (not shown in the drawings) for training the machine learning model.
For example, the model training unit may train the machine learning module by: at least one attribute information corresponding to the actual recommended content which is recommended in history is acquired, the characteristics of the training sample corresponding to the actual recommended content are generated at least based on the acquired at least one attribute information according to the same characteristic processing mode as the characteristics of the prediction sample, the result of whether the actual recommended content is clicked is used as a mark of the training sample, and a machine learning model is trained based on a set composed of a plurality of training samples.
Optionally, the system for recommending content according to an exemplary embodiment of the present invention may further include: and the model updating unit is used for updating the machine learning model.
For example, the model updating unit may update the machine learning model by: the real results of whether the candidate content is clicked after being recommended to the user facing the candidate content are recorded, so that the machine learning model is retrained based on at least one attribute information corresponding to the candidate content and the real results of whether the candidate content is clicked after being recommended to the user facing the candidate content.
The optimization index information acquisition unit 30 acquires optimization index information for measuring the degree to which the candidate content is worth being recommended.
Alternatively, the optimization index information may refer to a subsequent index directly related to the candidate content being clicked, for example, the optimization index information may include historical statistics representing actions that the candidate content is further accepted after being clicked and/or price information measuring the benefits obtained after the candidate content is recommended to the user.
In an alternative embodiment, the historical statistics may include the retention rate of candidate content counted over a predetermined period of time after being clicked. Here, the retention rate may refer to a probability that a user reserves a contact way for candidate contents after clicking the candidate contents.
Alternatively, the optimization index information acquisition unit 30 may acquire the retention rate of the candidate content from the provider of the candidate content, and/or predict the retention rate of the candidate content after being clicked for the candidate content.
The recommendation index determining unit 40 determines a recommendation index that recommends candidate contents to the user based on the estimated click rate of the candidate contents together with the optimization index information.
Alternatively, the recommendation index determining unit 40 may further normalize price information measuring the benefits obtained after the candidate content is recommended to the user. In this case, the recommendation index determining unit 40 may determine a recommendation index that recommends candidate contents to the user based on the estimated click rate and at least one of the history statistical information and the price information after normalization processing.
As an example, the recommendation index may correspond to a product of the estimated click rate and at least one of the historical statistics and the normalized price information.
Optionally, the system for recommending content according to an exemplary embodiment of the present invention may further include: a content recommendation unit (not shown in the figure) for recommending candidate contents to the user based on the recommendation index.
In one example, the content recommendation unit may recommend the candidate content having the highest recommendation index among the set of candidate contents to the user.
In another example, the content recommendation unit may create a list of recommended content, and provide the created list of recommended content to the user. Here, the recommended content in the list may be the top-ranked candidate content of the recommendation index among the set of candidate contents.
As an example, the set of candidate content may include all candidate content or at least a portion of candidate content recalled for the characteristics of the user at hand.
The units included in the system for recommending content according to the exemplary embodiment of the present invention may be respectively configured as software, hardware, firmware, or any combination of the above to perform a specific function. For example, these means may correspond to application specific integrated circuits, to pure software code, or to modules of software in combination with hardware. Furthermore, one or more functions implemented by these means may also be performed uniformly by components in a physical entity apparatus (e.g., a processor, a client, a server, or the like).
A computing device for recommending content is also presented in an exemplary embodiment of the present invention. The computing devices may be deployed in servers or clients, as well as on node devices in a distributed network environment. Further, the computing device may be a PC computer, tablet device, personal digital assistant, smart phone, web application, or other device capable of executing the above-described set of instructions.
Here, the computing device need not be a single computing device, but may be any device or collection of circuits capable of executing the above-described instructions (or instruction set) alone or in combination. The computing device may also be part of an integrated control system or system manager, or may be configured as a portable electronic device that interfaces with locally or remotely (e.g., via wireless transmission).
In the computing device, the processor may include a Central Processing Unit (CPU), a Graphics Processor (GPU), a programmable logic device, a special purpose processor system, a microcontroller, or a microprocessor. By way of example, and not limitation, processors may also include analog processors, digital processors, microprocessors, multi-core processors, processor arrays, network processors, and the like.
Some of the operations described in the method for recommending content according to the exemplary embodiment of the present invention may be implemented in software, some of the operations may be implemented in hardware, and furthermore, the operations may be implemented in a combination of software and hardware.
The processor may execute instructions or code stored in one of the storage components, wherein the storage component may also store data. Instructions and data may also be transmitted and received over a network via a network interface device, which may employ any known transmission protocol.
The memory component may be integrated with the processor, for example, RAM or flash memory disposed within an integrated circuit microprocessor or the like. Further, the storage component may comprise a stand-alone device, such as an external disk drive, a storage array, or any other storage device usable by a database system. The storage component and the processor may be operatively coupled or may communicate with each other, such as through an I/O port, network connection, etc., such that the processor is able to read files stored in the storage component.
In addition, the computing device may also include a video display (such as a liquid crystal display) and a user interaction interface (such as a keyboard, mouse, touch input device, etc.). All components of the computing device may be connected to each other via buses and/or networks.
Operations involved in a method for recommending content according to exemplary embodiments of the present invention may be described as various interconnected or coupled functional blocks or functional diagrams. However, these functional blocks or functional diagrams may be equally integrated into a single logic device or operate at non-exact boundaries.
For example, as described above, a system is provided that includes at least one computing device and at least one storage device storing instructions that, when executed by the at least one computing device, cause the at least one computing device to perform the steps of: generating a prediction sample corresponding to candidate content aiming at the candidate content; predicting the probability of clicking the candidate content after being recommended to the user facing the candidate content according to the prediction sample, wherein the probability is used as the estimated click rate of the candidate content; acquiring optimization index information for measuring the recommended degree of the candidate content; and determining a recommendation index that recommends the candidate content to the user based on the estimated click rate along with the optimization index information.
That is, the method for recommending content shown in fig. 1 may be performed by the above-described computing device. Since the method for recommending content has been described in detail in fig. 1 to 3, the content of this part is not repeated in the present invention.
Alternatively, the above-described system and computing device for recommending content may be integrated in a server of a publisher of the candidate content, for example, may be integrated in a server of a web page that publishes the candidate content, or may also be integrated in a server of an application that publishes the candidate content. In addition, the recommendation index may be integrated in a third party server to obtain a recommendation index for recommending the candidate content to the user, and then the publisher of the candidate content recommends the candidate content to the user according to the recommendation index.
It should be appreciated that the method for recommending content according to an exemplary embodiment of the present invention may be implemented by a program recorded on a computer-readable medium, for example, according to an exemplary embodiment of the present invention, a computer-readable medium storing instructions may be provided, wherein the instructions, when executed by at least one computing device, cause the at least one computing device to perform the method of: generating a prediction sample corresponding to candidate content aiming at the candidate content; predicting the probability of clicking the candidate content after being recommended to the user facing the candidate content according to the prediction sample, wherein the probability is used as the estimated click rate of the candidate content; acquiring optimization index information for measuring the recommended degree of the candidate content; and determining a recommendation index that recommends the candidate content to the user based on the estimated click rate along with the optimization index information.
The computer program in the above-described computer readable medium may be run in an environment deployed in a computer device such as a client, a host, a proxy device, a server, etc., and it should be noted that the computer program may also be used to perform additional steps other than the above-described steps or to perform more specific processes when the above-described steps are performed, and the contents of these additional steps and further processes have been described with reference to fig. 1 to 3, and will not be repeated here.
It should be noted that the system for recommending content according to an exemplary embodiment of the present invention may completely rely on the execution of a computer program to implement the corresponding functions, i.e., each device corresponds to each step in the functional architecture of the computer program, so that the entire system is called through a specific software package (e.g., lib library) to implement the corresponding functions.
On the other hand, the respective units included in the system for recommending content according to the exemplary embodiment of the present invention may also be implemented by hardware, software, firmware, middleware, microcode, or any combination thereof. When implemented in software, firmware, middleware or microcode, the program code or code segments to perform the corresponding operations may be stored in a computer-readable medium, such as a storage medium, so that the processor can perform the corresponding operations by reading and executing the corresponding program code or code segments.
The foregoing description of exemplary embodiments of the invention has been presented only to be understood as illustrative and not exhaustive, and the invention is not limited to the exemplary 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 invention. Therefore, the protection scope of the present invention shall be subject to the scope of the claims.
Claims (22)
1. A method for recommending content, comprising:
generating a prediction sample corresponding to candidate content aiming at the candidate content;
predicting the probability of clicking the candidate content after being recommended to the user facing the candidate content according to the prediction sample, wherein the probability is used as the estimated click rate of the candidate content;
obtaining optimization index information for measuring the recommended degree of the candidate content, wherein the optimization index information comprises historical statistical information representing the action of the candidate content which is further accepted after being clicked and/or price information for measuring the income obtained after the candidate content is recommended to a user; and
determining a recommendation index that recommends the candidate content to the user based on the estimated click rate along with the optimization index information;
Wherein the step of generating the prediction samples corresponding to the candidate content comprises:
acquiring at least one attribute information corresponding to the candidate content, wherein the at least one attribute information comprises attribute information related to an environment, and the attribute information related to the environment comprises at least one of the following items: the browser version of the candidate content is displayed, the category of the terminal equipment displaying the candidate content, the model number of the terminal equipment, weather, seasons and recent hot events are displayed;
a prediction sample corresponding to the candidate content is generated based at least on the at least one acquired attribute information.
2. The method of claim 1, wherein the at least one attribute information further comprises attribute information related to the candidate content and/or a user to whom it is directed.
3. The method of claim 1 or 2, wherein the historical statistics include a retention rate of the candidate content after being clicked counted for a predetermined period of time, wherein the retention rate refers to a probability that each user reserves a contact address for the candidate content after clicking the candidate content.
4. A method according to claim 3, wherein the retention rate is obtained by:
Obtaining a retention rate of the candidate content from a provider of the candidate content; and/or
And predicting the retention rate of the candidate content after being clicked according to the candidate content.
5. The method of claim 1 or 2, wherein the method further comprises: normalizing price information measuring benefits obtained after the candidate content is recommended to the user,
wherein determining a recommendation index that recommends the candidate content to the user based on the estimated click rate along with the optimization index information comprises:
a recommendation index for recommending the candidate content to the user is determined based on the estimated click rate and at least one of the historical statistics and normalized price information.
6. The method of claim 5, wherein the recommendation index corresponds to a product of the estimated click rate and at least one of the historical statistics and normalized price information.
7. The method of claim 1 or 2, wherein predicting the probability that the candidate content is clicked after being recommended to the user based on the prediction sample as the estimated click rate of the candidate content comprises:
And predicting the probability of the candidate content being clicked after being recommended to the user according to the prediction sample by using a machine learning model, wherein the probability is used as the estimated click rate of the candidate content.
8. The method of claim 7, wherein the machine learning model is trained by:
acquiring at least one attribute information corresponding to a real recommended content which is historically recommended;
generating features of the training samples corresponding to the real recommended content based on at least the acquired at least one attribute information in the same feature processing manner as the features of the prediction samples;
taking the result of whether the real recommended content is clicked or not as a mark of a training sample; and
the machine learning model is trained based on a set of a plurality of training samples.
9. The method of claim 1, wherein the method further comprises: recommending the candidate content to the user based on the recommendation index by one of:
recommending the candidate content with the highest recommendation index in the candidate content collection to a user;
creating a list of recommended contents, and providing the list to a user, wherein the recommended contents in the list are candidate contents with top recommendation indexes in a set of candidate contents.
10. The method of claim 9, wherein the set of candidate content includes all candidate content or at least a portion of candidate content recalled for the characteristics of the faced user.
11. A system for recommending content, comprising:
a prediction sample generation unit that generates a prediction sample corresponding to candidate content with respect to the candidate content;
the estimated click rate determining unit predicts the probability of the candidate content being clicked after being recommended to the user facing the candidate content according to the prediction sample, and takes the probability as the estimated click rate of the candidate content;
an optimization index information acquisition unit that acquires optimization index information for measuring a degree to which the candidate content is worth being recommended, the optimization index information including historical statistical information indicating an action of the candidate content that is further accepted after being clicked and/or price information measuring a return obtained after the candidate content is recommended to a user; and
a recommendation index determining unit that determines a recommendation index that recommends the candidate content to the user based on the estimated click rate together with the optimization index information;
wherein the prediction sample generation unit acquires at least one attribute information corresponding to the candidate content, generates a prediction sample corresponding to the candidate content based at least on the acquired at least one attribute information, wherein the at least one attribute information includes environment-related attribute information including at least one of: the browser version of the candidate content is displayed, the category of the terminal equipment displaying the candidate content, the model number of the terminal equipment, weather, seasons and recent hot events.
12. The system of claim 11, wherein the at least one attribute information further comprises attribute information related to the candidate content and/or the user to which it is directed.
13. The system of claim 11 or 12, wherein the historical statistics include a retention rate of the candidate content after being clicked counted for a predetermined period of time, wherein the retention rate refers to a probability that each user reserves a contact address for the candidate content after clicking the candidate content.
14. The system according to claim 13, wherein the optimization index information acquisition unit acquires the electricity retention rate by:
obtaining a retention rate of the candidate content from a provider of the candidate content; and/or
And predicting the retention rate of the candidate content after being clicked according to the candidate content.
15. The system according to claim 11 or 12, wherein the recommendation index determining unit further normalizes price information measuring a return obtained after the candidate content is recommended to a user, and determines a recommendation index recommending the candidate content to the user based on the estimated click rate and at least one of the history statistical information and the normalized price information.
16. The system of claim 15, wherein the recommendation index corresponds to a product of the estimated click rate and at least one of the historical statistics and normalized price information.
17. The system according to claim 11 or 12, wherein the estimated click rate determination unit predicts, as the estimated click rate of the candidate content, a probability that the candidate content is clicked after being recommended to the user, for the prediction sample using a machine learning model.
18. The system of claim 17, further comprising a model training unit that trains the machine learning model by:
acquiring at least one attribute information corresponding to a real recommended content which is historically recommended;
generating features of the training samples corresponding to the real recommended content based on at least the acquired at least one attribute information in the same feature processing manner as the features of the prediction samples;
taking the result of whether the real recommended content is clicked or not as a mark of a training sample; and
the machine learning model is trained based on a set of a plurality of training samples.
19. The system of claim 11, wherein the system further comprises: a content recommendation unit recommending the candidate content to the user based on the recommendation index by one of:
recommending the candidate content with the highest recommendation index in the candidate content collection to a user;
creating a list of recommended contents, and providing the list to a user, wherein the recommended contents in the list are candidate contents with top recommendation indexes in a set of candidate contents.
20. The system of claim 19, wherein the set of candidate content includes all candidate content or at least a portion of candidate content recalled for the characteristics of the user being faced.
21. A system comprising at least one computing device and at least one storage device storing instructions that, when executed by the at least one computing device, cause the at least one computing device to perform the method of any of claims 1-10.
22. A computer readable storage medium storing instructions which, when executed by at least one computing device, cause the at least one computing device to perform the method of any of claims 1 to 10.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811621052.5A CN111382346B (en) | 2018-12-28 | 2018-12-28 | Method and system for recommending content |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811621052.5A CN111382346B (en) | 2018-12-28 | 2018-12-28 | Method and system for recommending content |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111382346A CN111382346A (en) | 2020-07-07 |
| CN111382346B true CN111382346B (en) | 2023-09-01 |
Family
ID=71220452
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811621052.5A Active CN111382346B (en) | 2018-12-28 | 2018-12-28 | Method and system for recommending content |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111382346B (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104598511A (en) * | 2014-10-23 | 2015-05-06 | 腾讯科技（深圳）有限公司 | Method, device and system for recommending search results |
| CN104965890A (en) * | 2015-06-17 | 2015-10-07 | 深圳市腾讯计算机系统有限公司 | Advertisement recommendation method and apparatus |
| CN105095470A (en) * | 2015-08-03 | 2015-11-25 | 百度在线网络技术（北京）有限公司 | Data recommendation method and device for application program |
| CN107958030A (en) * | 2017-11-17 | 2018-04-24 | 北京奇虎科技有限公司 | Video front cover recommended models optimization method and device |
| CN108446382A (en) * | 2018-03-20 | 2018-08-24 | 百度在线网络技术（北京）有限公司 | Method and apparatus for pushed information |
| CN108510313A (en) * | 2018-03-07 | 2018-09-07 | 阿里巴巴集团控股有限公司 | A kind of prediction of information transferring rate, information recommendation method and device |
| CN108921300A (en) * | 2018-06-21 | 2018-11-30 | 第四范式（北京）技术有限公司 | The method and apparatus for executing automaton study |
| CN109062919A (en) * | 2018-05-31 | 2018-12-21 | 腾讯科技（深圳）有限公司 | A kind of content recommendation method and device based on deeply study |
| CN109063104A (en) * | 2018-07-27 | 2018-12-21 | 百度在线网络技术（北京）有限公司 | Method for refreshing, device, storage medium and the terminal device of recommendation information |
| CN109086439A (en) * | 2018-08-15 | 2018-12-25 | 腾讯科技（深圳）有限公司 | Information recommendation method and device |
-
2018
- 2018-12-28 CN CN201811621052.5A patent/CN111382346B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104598511A (en) * | 2014-10-23 | 2015-05-06 | 腾讯科技（深圳）有限公司 | Method, device and system for recommending search results |
| CN104965890A (en) * | 2015-06-17 | 2015-10-07 | 深圳市腾讯计算机系统有限公司 | Advertisement recommendation method and apparatus |
| CN105095470A (en) * | 2015-08-03 | 2015-11-25 | 百度在线网络技术（北京）有限公司 | Data recommendation method and device for application program |
| CN107958030A (en) * | 2017-11-17 | 2018-04-24 | 北京奇虎科技有限公司 | Video front cover recommended models optimization method and device |
| CN108510313A (en) * | 2018-03-07 | 2018-09-07 | 阿里巴巴集团控股有限公司 | A kind of prediction of information transferring rate, information recommendation method and device |
| CN108446382A (en) * | 2018-03-20 | 2018-08-24 | 百度在线网络技术（北京）有限公司 | Method and apparatus for pushed information |
| CN109062919A (en) * | 2018-05-31 | 2018-12-21 | 腾讯科技（深圳）有限公司 | A kind of content recommendation method and device based on deeply study |
| CN108921300A (en) * | 2018-06-21 | 2018-11-30 | 第四范式（北京）技术有限公司 | The method and apparatus for executing automaton study |
| CN109063104A (en) * | 2018-07-27 | 2018-12-21 | 百度在线网络技术（北京）有限公司 | Method for refreshing, device, storage medium and the terminal device of recommendation information |
| CN109086439A (en) * | 2018-08-15 | 2018-12-25 | 腾讯科技（深圳）有限公司 | Information recommendation method and device |
Non-Patent Citations (1)
| Title |
|---|
| 一种基于实时CTR的移动应用商店内容推荐改进算法;冯欣;夏旸;;长春理工大学学报(自然科学版)(第02期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111382346A (en) | 2020-07-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11551239B2 (en) | Characterizing and modifying user experience of computing environments based on behavior logs | |
| US11048530B1 (en) | Predictive action modeling to streamline user interface | |
| US11109083B2 (en) | Utilizing a deep generative model with task embedding for personalized targeting of digital content through multiple channels across client devices | |
| CN108230058B (en) | Product recommendation method and system | |
| WO2021135562A1 (en) | Feature validity evaluation method and apparatus, and electronic device and storage medium | |
| US11163783B2 (en) | Auto-selection of hierarchically-related near-term forecasting models | |
| CN111400613A (en) | Article recommendation method, device, medium and computer equipment | |
| US11809505B2 (en) | Method for pushing information, electronic device | |
| WO2023050143A1 (en) | Recommendation model training method and apparatus | |
| CN110717597A (en) | Method and device for acquiring time sequence characteristics by using machine learning model | |
| CN114741608A (en) | News recommendation method, device, equipment and storage medium based on user portrait | |
| CN111126649B (en) | Method and device for generating information | |
| CN112989174A (en) | Information recommendation method and device, medium and equipment | |
| WO2020211616A1 (en) | Method and device for processing user interaction information | |
| JP2022143937A (en) | Information processing apparatus, information processing method, and information processing program | |
| CN117194779A (en) | Marketing system optimization method, device and equipment based on artificial intelligence | |
| CN116703466A (en) | System access quantity prediction method based on improved wolf algorithm and related equipment thereof | |
| CN117217812A (en) | User behavior prediction method and device, computer equipment and storage medium | |
| JP2021189983A (en) | Device, system, program, and method for predicting purchase trend | |
| CN111382346B (en) | Method and system for recommending content | |
| US20230153843A1 (en) | System to combine intelligence from multiple sources that use disparate data sets | |
| CN112818231A (en) | Information delivery method and device, electronic equipment and storage medium | |
| CN110706772A (en) | Sorting method and device, electronic equipment and storage medium | |
| US20240152955A1 (en) | Position-bias correction for predictive and ranking systems | |
| US12038823B2 (en) | Hierarchical attention time-series (HAT) model for behavior prediction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |