CN116801056A - Video recommendation method and device - Google Patents
Video recommendation method and device Download PDFInfo
- Publication number
- CN116801056A CN116801056A CN202210256032.2A CN202210256032A CN116801056A CN 116801056 A CN116801056 A CN 116801056A CN 202210256032 A CN202210256032 A CN 202210256032A CN 116801056 A CN116801056 A CN 116801056A
- Authority
- CN
- China
- Prior art keywords
- video
- current user
- current
- unviewed
- probability vector
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 55
- 239000013598 vector Substances 0.000 claims abstract description 109
- 230000007704 transition Effects 0.000 claims abstract description 68
- 239000011159 matrix material Substances 0.000 claims abstract description 46
- 230000006399 behavior Effects 0.000 claims description 14
- 238000007637 random forest analysis Methods 0.000 claims description 13
- 238000004590 computer program Methods 0.000 claims description 12
- 238000004422 calculation algorithm Methods 0.000 claims description 9
- 239000000243 solution Substances 0.000 description 9
- 238000004891 communication Methods 0.000 description 6
- 230000008569 process Effects 0.000 description 6
- 230000008859 change Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 239000002585 base Substances 0.000 description 1
- 239000003637 basic solution Substances 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000002194 synthesizing effect Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
- H04N21/466—Learning process for intelligent management, e.g. learning user preferences for recommending movies
- H04N21/4662—Learning process for intelligent management, e.g. learning user preferences for recommending movies characterized by learning algorithms
- H04N21/4663—Learning process for intelligent management, e.g. learning user preferences for recommending movies characterized by learning algorithms involving probabilistic networks, e.g. Bayesian networks
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
- H04N21/466—Learning process for intelligent management, e.g. learning user preferences for recommending movies
- H04N21/4668—Learning process for intelligent management, e.g. learning user preferences for recommending movies for recommending content, e.g. movies
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a video recommendation method, which comprises the following steps: acquiring an initial state probability vector of a current user and a transition probability matrix of the current user from a current video to an unviewed video; determining a state probability vector for the current user to transition from the current video to each unviewed video based on the initial state probability vector and the transition probability matrix; predicting the video to be recommended of the current user based on all the state probability vectors and the category labels of the current user. According to the video recommending method and device, the state probability vector is determined based on the initial state probability vector and the transition probability matrix, and the video to be recommended of the current user is predicted according to the class label and the state probability vector of the current user, so that the computing efficiency can be improved, the video can be recommended to the user more accurately, and the preference of the user on the video in real time can be met more accurately.
Description
Technical Field
The present invention relates to the field of computer technologies, and in particular, to a video recommendation method and apparatus.
Background
With the development of internet technology, more and more users like to watch videos on a network, and as the videos can efficiently transmit image and sound information, the videos meet the visual sense and the auditory sense of people, can play a great role in education, art, social contact and entertainment, and how to recommend proper videos to users according to the preference or the requirement of the users becomes a research-worthy problem.
The existing method for recommending videos to users has relatively fixed rules, single recommendation mechanism and poor accuracy of recommendation results, and is difficult to accord with the preference of users in real-time change.
Disclosure of Invention
The invention provides a video recommending method and device, which are used for solving the defects that in the prior art, a method for recommending videos to users is relatively fixed in rule, single in recommending mechanism and poor in recommending result accuracy, and is difficult to accord with the preference of the users in real time, so that the computing efficiency is improved, the videos are recommended to the users more accurately, and the preference of the users on the video in real time can be met more accurately.
The invention provides a video recommendation method, which comprises the following steps: acquiring an initial state probability vector of a current user and a transition probability matrix of the current user from a current video to an unviewed video; determining a state probability vector for the current user to transition from the current video to each unviewed video based on the initial state probability vector and the transition probability matrix; predicting the video to be recommended of the current user based on all the state probability vectors and the category labels of the current user.
According to the video recommendation method provided by the invention, the obtaining the transition probability matrix of the current user from the current video to the unviewed video comprises the following steps: acquiring the preference state of the current user for the current video, the preference state of other users for the current video and the preference state of other users for the unviewed video; and determining a transition probability matrix for the current user to transition from the current video to the unviewed video based on the preference state of the current user for the current video, the preference state of other users for the current video and the preference state of other users for the unviewed video.
According to the video recommendation method provided by the invention, the determining the state probability vector of the current user from the current video to each unviewed video based on the initial state probability vector and the transition probability matrix comprises the following steps: constructing a Markov learning chain based on the initial state probability vector and the transition probability matrix; based on the Markov learning chain, a state probability vector for the current user to transition from the current video to each unviewed video is determined.
According to the video recommendation method provided by the invention, the step of acquiring the category label of the current user comprises the following steps: acquiring reference behavior information of a plurality of users, and counting the number of discrete variables and the number of continuous variables in the reference behavior information; determining derived features and variable class labels based on the number of discrete variables; constructing a random forest model based on the continuous variable, the discrete variable, the derivative feature and the variable class label; and inputting the reference behavior information into the random forest model to obtain the variable class labels output by the random forest model.
According to the video recommendation method provided by the invention, the determining of the derived feature and the variable category label based on the number of the discrete variables comprises the following steps: performing derivative treatment on the discrete variables larger than the discrete value quantity threshold value to obtain derivative characteristics; and clustering the discrete variables smaller than the discrete value quantity threshold to obtain the variable class labels.
According to the video recommendation method provided by the invention, the obtaining of the initial state probability vector of the current user comprises the following steps: acquiring historical browsing data of the current user; the initial state probability vector is determined based on a naive bayes algorithm.
According to the video recommendation method provided by the invention, based on the category label of the current user and the state probability vector, the video to be recommended of the current user is predicted, and the method comprises the following steps: acquiring a video category label of each unviewed video; and determining the video to be recommended from a plurality of unviewed videos based on the category label of the current user, the state probability vector and the video category label.
According to the video recommendation method provided by the invention, the video category labels of each unviewed video are obtained, and the method comprises the following steps: extracting subtitle information from a plurality of unviewed videos; clustering the caption information to obtain a reference label, and constructing a reference label data set based on the reference label; the video category labels for each unviewed video are determined based on association rules of the reference label dataset.
The invention also provides a video recommending device, which comprises: the first acquisition module is used for acquiring the initial state probability vector of the current user; the second acquisition module is used for acquiring a transition probability matrix of the current user from the current video to the unviewed video; a determining module configured to determine a state probability vector for the current user to transition from the current video to each unviewed video based on the initial state probability vector and the transition probability matrix; and the prediction module is used for predicting the video to be recommended of the current user based on the category label of the current user and the state probability vector.
The invention also provides an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the video recommendation method as described above when executing the program.
The present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements a video recommendation method as described in any of the above.
The invention also provides a computer program product comprising a computer program which when executed by a processor implements a video recommendation method as described in any one of the above.
According to the video recommending method and device, the category label of the current user is determined according to the target feature information of the current user, the state probability vector is determined based on the initial state probability vector and the transition probability matrix, and the video to be recommended of the current user is predicted according to the category label and the state probability vector of the current user, so that the computing efficiency can be improved, the video can be recommended to the user more accurately, and the preference of the user on the real-time change of the video can be met more accurately.
Drawings
In order to more clearly illustrate the invention or the technical solutions of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic flow chart of a video recommendation method provided by the invention;
fig. 2 is a schematic structural diagram of a video recommendation device provided by the present invention;
fig. 3 is a schematic structural diagram of an electronic device provided by the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is apparent that the described embodiments are some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The video recommendation method and apparatus of the present invention are described below with reference to fig. 1 to 3.
As shown in fig. 1, the present invention provides a video recommendation method, which includes the following steps 110 to 130.
Step 110, obtaining an initial state probability vector of a current user and a transition probability matrix of the current user from a current video to an unviewed video;
it can be appreciated that the video recommendation method can be applied to an electronic device, and the electronic device can be a mobile phone, a tablet computer, a notebook computer or a desktop computer. The current user refers to an object of video recommendation, and when the current user uses the electronic device, the video recommendation method of the embodiment can be applied to recommend videos to the current user.
The initial state probability vector may include: the favorite state and the disfavorite state, the initial state probability vector is expressed in the form of a vector, the initial state probability vector can be extracted based on historical browsing data of the current user, and the initial state probability vector can comprise the favorite state of the current user on the current video.
Meanwhile, the target feature information of the current user can be acquired first, and the category label of the current user can be determined based on the target feature information. The target feature information of the current user may be a user's viewing or clicking feature of the video, i.e., a feature variable based on user behavior formed from the user's access to the video. For example, when the video is an educational video, the target feature information of the user may include: user base attributes such as child age, class, grade, school, and gender; the target feature information of the user may further include: such as purchasing courses, the number of course comments, the time of watching, the number of watching, collecting courses, praise of courses and the like.
Here, the category label of the current user can be determined according to the target feature information of the current user, that is, the current user is marked according to the operation behavior of the current user in the history browsing process, and which group the current user belongs to is judged, so that the variable category label of the user can be realized.
The transition probability matrix of the current user from the current video to the unviewed video refers to the transition probability of the current user from the unviewed video after watching the current video, for example, the current user must slide to the B video after watching the a video, and then the transition probability of the current user from the a video to the B video is 100%. Here, a transition probability matrix of the current user for a plurality of videos may be obtained, where the plurality of videos may be videos that the user has not yet watched.
Step 120, determining a state probability vector for the current user to transition from the current video to each unviewed video based on the initial state probability vector and the transition probability matrix.
It may be appreciated that the state probability vector for the current user to transition from the viewed current video to each unviewed video may be calculated based on the initial state probability vector and the transition probability matrix, where the state probability vector may be a liked state probability of the unviewed video by the current user.
The initial state probability vector may represent a probability vector at the start of a process of predicting the occurrence probability of an event. Meanwhile, the probability of each two events in the process of predicting the occurrence probability of the event can be mutually shifted under a certain condition, so that a shift probability matrix can represent the shift relation between the occurrence probabilities of the two events, and a state probability vector refers to a probability vector of occurrence of a target event to be predicted and can be also called a probability vector of a target moment in the process of prediction.
And 130, predicting the video to be recommended of the current user based on all the state probability vectors and the category labels of the current user.
It can be understood that intelligent combination processing can be performed according to the category label and all the state probability vectors of the current user, a target data set for viewing the video behavior portraits by the user can be constructed, the category label and all the state probability vectors of the current user are put into the target data set in a parallel manner, the video to be recommended of the current user is predicted according to the target data set, namely, the current user portraits are drawn according to the category label and all the state probability vectors of the current user, and the corresponding video to be recommended is matched with the current user according to the drawn user portraits.
The category labels of the current user and all state probability vectors are placed in a target data set in a parallel mode, the video to be recommended of the current user is predicted according to the target data set, the video can be dynamically decided to be recommended to the user, the experience of the user is improved, the viscosity of the user is enhanced, the problems of low data sparsity, low operation efficiency and low accuracy in the existing recommendation method are solved, and the video can be intelligently recommended to the user.
According to the video recommendation method provided by the invention, the category label of the current user is determined according to the target feature information of the current user, the state probability vector is determined based on the initial state probability vector and the transition probability matrix, and the video to be recommended of the current user is predicted according to the category label and the state probability vector of the current user, so that the operation efficiency can be improved, the video can be recommended to the user more accurately, and the preference of the user on the real-time change of the video can be met more accurately.
In some embodiments, the obtaining the transition probability matrix of the current user from the current video to each video in step 110 includes: acquiring the preference state of a current user for a current video, the preference state of other users for the current video and the preference state of other users for unviewed video; and determining a transition probability matrix for the current user to transition from the current video to the unviewed video based on the preference state of the current user for the current video, the preference state of other users for the current video and the preference state of other users for the unviewed video.
It can be understood that the transition probability matrix of the current user from the watched video to the unviewed video can be calculated according to the video watched by the current user and the preference states of all users for the video watched by the current user.
The state probability vector may be a preference state probability. The state probability vector of the current user for transferring from the watched video to each unviewed video can be calculated based on the initial state probability vector and the transfer probability matrix; wherein the state probability vector includes a preference state probability of the user for the unviewed video.
In some embodiments, determining a state probability vector for a current user to transition from a current video to each unviewed video based on the initial state probability vector and the transition probability matrix comprises: constructing a Markov learning chain based on the initial state probability vector and the transition probability matrix; based on the Markov learning chain, a state probability vector for the current user to transition from the current video to each unviewed video is determined.
It will be appreciated that after the initial state probability vector and the transition probability matrix are obtained, a Markov learning chain may be constructed from the initial state probability vector and the transition probability matrix. A markov learning chain is an abstract model used to describe a series of interrelated events or states. In other words, the probability distribution of which event (switch to which state) will occur next is known on the premise that each event or state has already occurred.
The markov learning chain essentially consists of transition probability distribution meeting the markov property, and a formula for calculating a state probability vector from an initial state probability vector and a transition probability matrix through the markov learning chain can be:
P ss′ ＝P(S t+1 ＝s′|S t ＝s)；
wherein S and S' are both time state sequences, S t Representing the current state S t+1 Representing the next state, and P represents the probability of a state occurring.
Here, after constructing the markov learning chain from the initial state probability vector and the transition probability matrix, the state probability vector for the current user to transition from the current video to each unviewed video may be determined based on the markov learning chain.
In some embodiments, determining the category label of the current user based on the target feature information comprises: acquiring reference behavior information of a plurality of users, and counting the number of discrete variables and the number of continuous variables in the reference behavior information; determining derived features and variable class labels based on the number of discrete variables; constructing a random forest model based on continuous variables, discrete variables, derived features and variable class labels; and inputting the reference behavior information into the random forest model to obtain a variable class label output by the random forest model.
It can be understood that the characteristic variables obtained by the reference behavior information according to the counting mode are classified into discrete variables, other types are classified into continuous variables, and if the variation range of the variable values is small for the discrete variables, one variable value corresponds to one group, and single item grouping is carried out; if the variable value has a large fluctuation range and a large number of variable values, the whole variable value is divided into several sections in turn, the sections to be merged are determined according to the size of each variable value, the distance between the sections is called group distance, and group distance type grouping is performed.
In some embodiments, determining derived features and variable class labels based on the number of discrete variables includes: performing derivative treatment on discrete variables larger than the threshold value of the number of discrete values to obtain derivative characteristics; and clustering discrete variable which is smaller than the discrete value quantity threshold value to obtain variable class labels.
The discrete variable exceeding the set discrete value quantity threshold can be subjected to derivative treatment according to the quantity of the discrete variable, and derivative characteristics are established; and clustering discrete variables which do not exceed the discrete value quantity threshold by adopting K-Modes, and outputting class labels.
The random forest model can calculate and evaluate the importance of each factor feature through a feature classification process at the same time of prediction. The calculation of the feature influence force needs to be carried out by means of the Gini index at the node splitting time, and the formula is as follows:
Ii(A)＝Gini(Di)-Gini(Di,A)Ii(A)＝Gini(Di)-Gini(Di,A)；
S(A)＝∑iIi(A)；
and the Ii (A) represents the descending value of the Gini index relative to the parent node before splitting after the node i is split into two child nodes according to the feature A, so that the absolute importance S (A) of the feature A can be defined as the sum of Ii (A) at all nodes split according to the feature A, and the absolute importance of all the features is normalized to obtain the importance score of each feature.
Constructing a random forest model according to continuous variables, discrete variables, derivative features and variable class labels; and inputting the reference behavior information into the random forest model to obtain variable category labels output by the random forest model, so as to realize crowd classification and realize recommendation of different videos for users of different categories of crowds.
In some embodiments, obtaining the initial state probability vector for the current user includes: acquiring historical browsing data of a current user; the initial state probability vector is determined based on a naive bayes algorithm.
It can be appreciated that the initial state probability vector in the historical browsing data of the user can be calculated based on a naive bayes algorithm; the initial state probability vector comprises a like state and a dislike state, and the specific calculation process is as follows:
p (a|b) represents the probability of event a occurring on the premise that event B has occurred, and is called the conditional probability of event a occurring under event B. The basic solution formula is as follows:
p (a|b) can be derived directly, but P (b|a) is difficult to derive directly, but P (b|a) is of more concern here, and the naive bayes algorithm can open the way to obtain P (b|a) from P (a|b), with the following formula:
thus, here, the preference state of the current user for the current video may be determined based on a naive bayes algorithm, that is, the initial state probability vector is obtained.
In the application scenario of the present embodiment, the naive bayes algorithm is assumed that each feature is independent, and a formula of the naive bayes algorithm may be expressed as:
specifically, the historical browsing data of the current user may have a large number of features, and in order to obtain the preference state P of the current user for the current video (like the feature related to the current video), only the judgment of P (like the feature related to the current video) P (like the current video) and P (like the feature related to the current video) is needed, so that the preference state of the current user for the current video, that is, the initial state probability vector, can be obtained according to the above formula.
In some embodiments, predicting the video to be recommended for the current user based on the category labels and the state probability vectors for the current user includes: acquiring a video category label of each unviewed video; and determining the video to be recommended from the plurality of unviewed videos based on the category label, the state probability vector and the video category label of the current user.
It can be understood that the corresponding relation between the video and the video category label can be established, that is, each video can be marked, and each video can have the video category label corresponding to the video, so that after the category label and the state probability vector of the current user are obtained, a user portrait dataset can be established according to the category label and the state probability vector of the current user, the user portrait dataset is matched with the video category label, and the video corresponding to the video category label is found, so that the video to be recommended can be accurately determined from a plurality of videos.
In some embodiments, obtaining a video category label for each unviewed video includes: extracting subtitle information from a plurality of unviewed videos; clustering the caption information to obtain a reference label, and constructing a reference label data set based on the reference label; video category tags for each unviewed video are determined based on association rules of the reference tag dataset.
It can be understood that after the video is obtained, the video can be preprocessed, and the video caption in the video is converted into a standard binary image to be identified; extracting the size, length and width, stroke type and stroke density of the video, then regarding the text in the video as a special symbol with characteristics, synthesizing the time domain characteristic information of the video, and further converting the time domain characteristic information into a binary image to be identified.
Meanwhile, dividing the video into independent shots, positioning caption areas in the shot segments by utilizing the difference of two successive frames caused by the appearance and disappearance of captions, obtaining a series of rectangular areas with characters in an original image, separating the rectangular areas to obtain a series of sub-images of the original image, and removing the background from a character block to obtain a binary image only containing character information; and extracting the pictures and the binarized images in the voice and picture sample library to be recognized, and extracting subtitle information.
Clustering the caption information to obtain reference labels, constructing a reference label data set based on the reference labels, determining association rules among the reference labels through the reference label data set, and mining the association rules among the reference labels by using an Apriori algorithm so as to construct a reference label system.
For example, all reference tags may be found, and reference tag data sets may be created that occur more frequently than or equal to a predetermined minimum support. By referring to the tag data set, the strong association rule generated by the reference tag data set is found, and then the association rule between the reference tags is found, for example, when the video is educational material, the reference tags may include: discipline, applicable age group, adaptive textbook or resource quality rating, etc.
After deriving the association rules between the reference tags, a video category tag for each unviewed video may be determined based on the association rules of the reference tag dataset.
The video recommendation device provided by the invention is described below, and the video recommendation device described below and the video recommendation method described above can be referred to correspondingly.
As shown in fig. 2, the present invention further provides a video recommendation device, which includes: the acquisition module 210, the determination module 220, and the prediction module 230.
The obtaining module 210 is configured to obtain an initial state probability vector of a current user, and a transition probability matrix of the current user transitioning from a current video to an unviewed video.
A determining module 230 is configured to determine a state probability vector for the current user to transition from the current video to each unviewed video based on the initial state probability vector and the transition probability matrix.
The prediction module 240 is configured to predict the video to be recommended for the current user based on all the state probability vectors and the category labels of the current user.
According to the video recommending device provided by the invention, the category label of the current user is determined according to the target characteristic information of the current user, the state probability vector is determined based on the initial state probability vector and the transition probability matrix, and the video to be recommended of the current user is predicted according to the category label and the state probability vector of the current user, so that the computing efficiency can be improved, the video can be recommended to the user more accurately, and the preference of the user on the real-time change of the video can be met more accurately.
Fig. 3 illustrates a physical schematic diagram of an electronic device, as shown in fig. 3, where the electronic device may include: processor 310, communication interface (Communications Interface) 320, memory 330 and communication bus 340, wherein processor 310, communication interface 320, memory 330 accomplish communication with each other through communication bus 340. The processor 310 may invoke logic instructions in the memory 330 to perform a video recommendation method comprising: acquiring an initial state probability vector of a current user and a transition probability matrix of the current user from a current video to an unviewed video; determining a state probability vector for a current user to transition from the current video to each unviewed video based on the initial state probability vector and the transition probability matrix; predicting the video to be recommended of the current user based on all the state probability vectors and the category labels of the current user.
Further, the logic instructions in the memory 330 described above may be implemented in the form of software functional units and may be stored in a computer-readable storage medium when sold or used as a stand-alone product. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
In another aspect, the present invention also provides a computer program product, the computer program product comprising a computer program, the computer program being storable on a non-transitory computer readable storage medium, the computer program, when executed by a processor, being capable of executing the video recommendation method provided by the above methods, the method comprising: acquiring an initial state probability vector of a current user and a transition probability matrix of the current user from a current video to an unviewed video; determining a state probability vector for a current user to transition from the current video to each unviewed video based on the initial state probability vector and the transition probability matrix; predicting the video to be recommended of the current user based on all the state probability vectors and the category labels of the current user.
In yet another aspect, the present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, is implemented to perform the video recommendation method provided by the above methods, the method comprising: acquiring an initial state probability vector of a current user and a transition probability matrix of the current user from a current video to an unviewed video; determining a state probability vector for a current user to transition from the current video to each unviewed video based on the initial state probability vector and the transition probability matrix; predicting the video to be recommended of the current user based on all the state probability vectors and the category labels of the current user.
The apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course may be implemented by means of hardware. Based on this understanding, the foregoing technical solution may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the respective embodiments or some parts of the embodiments.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (11)
1. A video recommendation method, comprising:
acquiring an initial state probability vector of a current user and a transition probability matrix of the current user from a current video to an unviewed video;
determining a state probability vector for the current user to transition from the current video to each unviewed video based on the initial state probability vector and the transition probability matrix;
predicting the video to be recommended of the current user based on all the state probability vectors and the category labels of the current user.
2. The video recommendation method according to claim 1, wherein the obtaining a transition probability matrix for the current user to transition from a current video to an unviewed video comprises:
acquiring the preference state of the current user for the current video, the preference state of other users for the current video and the preference state of other users for the unviewed video;
and determining a transition probability matrix for the current user to transition from the current video to the unviewed video based on the preference state of the current user for the current video, the preference state of other users for the current video and the preference state of other users for the unviewed video.
3. The video recommendation method according to claim 1, wherein said determining a state probability vector for the current user to transition from the current video to each unviewed video based on the initial state probability vector and the transition probability matrix comprises:
constructing a Markov learning chain based on the initial state probability vector and the transition probability matrix;
based on the Markov learning chain, a state probability vector for the current user to transition from the current video to each unviewed video is determined.
4. The video recommendation method according to claim 1, wherein the step of acquiring the category label of the current user comprises:
acquiring reference behavior information of a plurality of users, and counting the number of discrete variables and the number of continuous variables in the reference behavior information;
determining derived features and variable class labels based on the number of discrete variables;
constructing a random forest model based on the continuous variable, the discrete variable, the derivative feature and the variable class label;
and inputting the reference behavior information into the random forest model to obtain the variable class labels output by the random forest model.
5. The video recommendation method according to claim 4, wherein said determining derived features and variable category labels based on the number of discrete variables comprises:
performing derivative treatment on the discrete variables larger than the discrete value quantity threshold value to obtain derivative characteristics;
and clustering the discrete variables smaller than the discrete value quantity threshold to obtain the variable class labels.
6. The video recommendation method according to any one of claims 1 to 5, wherein said obtaining an initial state probability vector for the current user comprises:
acquiring historical browsing data of the current user;
the initial state probability vector is determined based on a naive bayes algorithm.
7. The video recommendation method according to any one of claims 1 to 5, wherein said predicting a video to be recommended for a current user based on a category label of the current user and the state probability vector comprises:
acquiring a video category label of each unviewed video;
and determining the video to be recommended from a plurality of unviewed videos based on the category label of the current user, the state probability vector and the video category label.
8. The video recommendation method of claim 7, wherein said obtaining a video category label for each unviewed video comprises:
extracting subtitle information from a plurality of unviewed videos;
clustering the caption information to obtain a reference label, and constructing a reference label data set based on the reference label;
the video category labels for each unviewed video are determined based on association rules of the reference label dataset.
9. A video recommendation device, comprising:
the acquisition module is used for acquiring an initial state probability vector of a current user and a transition probability matrix of the current user from a current video to an unviewed video;
a determining module configured to determine a state probability vector for the current user to transition from the current video to each unviewed video based on the initial state probability vector and the transition probability matrix;
and the prediction module is used for predicting the video to be recommended of the current user based on the category label of the current user and the state probability vector.
10. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the video recommendation method of any one of claims 1 to 8 when the program is executed by the processor.
11. A non-transitory computer readable storage medium having stored thereon a computer program, which when executed by a processor implements the video recommendation method according to any of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210256032.2A CN116801056A (en) | 2022-03-15 | 2022-03-15 | Video recommendation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210256032.2A CN116801056A (en) | 2022-03-15 | 2022-03-15 | Video recommendation method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116801056A true CN116801056A (en) | 2023-09-22 |
Family
ID=88042360
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210256032.2A Pending CN116801056A (en) | 2022-03-15 | 2022-03-15 | Video recommendation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116801056A (en) |
-
2022
- 2022-03-15 CN CN202210256032.2A patent/CN116801056A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106973244B (en) | Method and system for automatically generating image captions using weak supervision data | |
| CN108763362B (en) | Local model weighted fusion Top-N movie recommendation method based on random anchor point pair selection | |
| CN106599226B (en) | Content recommendation method and content recommendation system | |
| CN112364204B (en) | Video searching method, device, computer equipment and storage medium | |
| US20200026756A1 (en) | Determining personality profiles based on online social speech | |
| CN113705299A (en) | Video identification method and device and storage medium | |
| CN111625715B (en) | Information extraction method and device, electronic equipment and storage medium | |
| CN113761253A (en) | Video tag determination method, device, equipment and storage medium | |
| CN112257452A (en) | Emotion recognition model training method, device, equipment and storage medium | |
| CN111680190A (en) | Video thumbnail recommendation method fusing visual semantic information | |
| CN113987274A (en) | Video semantic representation method and device, electronic equipment and storage medium | |
| CN113095901B (en) | Recommendation method, training method of related model, electronic equipment and storage device | |
| Sreeja et al. | A unified model for egocentric video summarization: an instance-based approach | |
| CN113032676B (en) | Recommendation method and system based on micro-feedback | |
| CN114330514A (en) | Data reconstruction method and system based on depth features and gradient information | |
| CN115640449A (en) | Media object recommendation method and device, computer equipment and storage medium | |
| CN115730073A (en) | Text processing method, device and storage medium | |
| Hanafi et al. | Word Sequential Using Deep LSTM and Matrix Factorization to Handle Rating Sparse Data for E‐Commerce Recommender System | |
| CN113569091B (en) | Video data processing method and device | |
| CN116801056A (en) | Video recommendation method and device | |
| Pandiaraj et al. | A Comparative Study on Emotion AI using Machine Learning and Deep Learning Models | |
| CN113987262A (en) | Video recommendation information determination method and device, electronic equipment and storage medium | |
| CN115130453A (en) | Interactive information generation method and device | |
| CN116704423B (en) | Hierarchical video character social interaction identification method, system, equipment and medium | |
| CN116150428B (en) | Video tag acquisition method and device, electronic equipment and storage medium |
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 |