CN115878897A - Video recommendation method and device, electronic equipment and storage medium - Google Patents
Video recommendation method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN115878897A CN115878897A CN202211599774.1A CN202211599774A CN115878897A CN 115878897 A CN115878897 A CN 115878897A CN 202211599774 A CN202211599774 A CN 202211599774A CN 115878897 A CN115878897 A CN 115878897A
- Authority
- CN
- China
- Prior art keywords
- video
- user
- recommended
- album
- cold
- 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 59
- 238000004364 calculation method Methods 0.000 claims abstract description 28
- 230000003068 static effect Effects 0.000 claims description 48
- 238000013145 classification model Methods 0.000 claims description 37
- 238000004590 computer program Methods 0.000 claims description 18
- 238000012549 training Methods 0.000 claims description 18
- 238000000605 extraction Methods 0.000 claims description 13
- 238000012545 processing Methods 0.000 claims description 12
- 238000010606 normalization Methods 0.000 claims description 7
- 102100035353 Cyclin-dependent kinase 2-associated protein 1 Human genes 0.000 description 55
- 238000010586 diagram Methods 0.000 description 14
- 230000008569 process Effects 0.000 description 9
- 230000006870 function Effects 0.000 description 8
- 239000011159 matrix material Substances 0.000 description 7
- 238000004422 calculation algorithm Methods 0.000 description 5
- 230000009471 action Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
Images
Abstract
The embodiment of the invention provides a video recommendation method, a video recommendation device, electronic equipment and a storage medium, wherein the method comprises the following steps: acquiring the watching information of each video album in the video album set and the user information of a cold start user, and clustering the video albums in the video album set according to the watching information of each video album to obtain a plurality of categories of video albums; selecting a video album to be recommended in each category aiming at the video albums in each category; calculating the matching degree between the video albums to be recommended and the cold start user according to the user information and the watching information of each video album to be recommended; and determining a target video album recommended to the cold-start user according to the matching degree. The video albums recommended to the cold start users are determined by clustering the video albums and combining the user information of the cold start users and the watching information of the video albums to be recommended, online calculation is not needed, and the video albums meeting the requirements of the cold start users can be quickly and effectively recommended to the cold start users.
Description
Technical Field
The present invention relates to the field of internet technologies, and in particular, to a video recommendation method, a video recommendation apparatus, an electronic device, and a computer-readable storage medium.
Background
With the continuous development of the intelligent terminal, videos can be watched through the video client on the intelligent terminal, wherein the cold start of the user plays an important role in video recommendation and is the last resort of service growth.
Video albums can be recommended to users who need cold start, typically based on EE (development and Exploration) algorithms, mining new content that may be of interest to the cold start user while recommending content of interest to the cold start user. When information generated by a cold-start user at a video client is acquired, the method needs online calculation to update a recommendation model in real time. However, the online computation has a high performance requirement on the model, and when the online computation has too much sample data, the model convergence speed is slow, so that the update speed of the model is reduced, the operating efficiency of the server is affected, and rapid and effective recommendation for the user is difficult to perform.
Disclosure of Invention
In view of the above problems, embodiments of the present invention are provided to provide a video recommendation method, a video recommendation apparatus, an electronic device, and a computer-readable storage medium that overcome or at least partially solve the above problems.
In order to solve the above problem, an embodiment of the present invention discloses a video recommendation method, including:
acquiring the watching information of each video album in the video album set and the user information of a cold start user;
clustering the video albums in the video album set according to the watching information of each video album to obtain a plurality of categories of video albums;
selecting a video album to be recommended of each category aiming at the video albums of each category;
calculating the matching degree between the video album to be recommended and the cold start user according to the user information and the watching information of each video album to be recommended;
and determining a target video album recommended to the cold-start user according to the matching degree.
Optionally, the selecting, for the video album of each category, a video album to be recommended in the category includes:
acquiring a clustering center corresponding to each category of video albums;
calculating the distance between each clustering center and each video album;
and aiming at each cluster center, selecting a plurality of video albums to be recommended in each category according to a preset selection number according to the sequence of the distance between each cluster center and each video album from near to far.
Optionally, the calculating, according to the user information and the viewing information of each video album to be recommended, a matching degree between the video album to be recommended and the cold-start user includes:
calculating the weight of the category of each video album to be recommended according to the distance between each video album to be recommended and each clustering center;
performing feature extraction on the user information of the cold start user to obtain the user static feature of the cold start user; the user static characteristics comprise a VIP identity characteristic, an age characteristic, an occupation characteristic, a preference characteristic and a city characteristic;
inputting the user static characteristics of the cold start user into a pre-trained classification model for processing to obtain the video preference category of the cold start user;
and calculating the matching degree between the video album to be recommended and the cold-start user based on the weight of the video favorite category of the cold-start user and the category of the video album to be recommended.
Optionally, the classification model is trained by:
acquiring training data, wherein the training data comprises user information of a sample user and video preference categories labeled by the sample user;
performing feature extraction on the user information of the sample user to obtain user static features of the sample user;
and training a classification model by adopting the user static characteristics of the sample user and the video preference categories marked by the sample user to obtain the classification model for identifying the video preference categories.
Optionally, the calculating, according to the distance between each video album to be recommended and each cluster center, the weight of the category to which the video album to be recommended belongs includes:
and carrying out normalization calculation on the reciprocal of the distance between each video album to be recommended and each clustering center to obtain the weight of the category to which the video album to be recommended belongs.
Optionally, the calculating, based on the weights of the video preference category of the cold-start user and the category to which the video album to be recommended belongs, a matching degree between each video album to be recommended and the cold-start user includes:
performing similarity calculation on the video favorite category of the cold start user and the category of the video album to be recommended by adopting a similarity calculation method to obtain the similarity between the video album to be recommended and the cold start user;
and performing weighted average calculation on the similarity by adopting the weight of the category to which the video album to be recommended belongs to obtain the matching degree between the video album to be recommended and the cold-start user.
Optionally, the clustering, according to the viewing information of each video album, the video albums in the video album set to obtain video albums of multiple categories includes:
performing feature extraction on the viewing information of each video album to obtain the viewing features of each video album;
and clustering the video albums according to the watching characteristics of the video albums to obtain video albums of multiple categories.
Optionally, before the acquiring the viewing information of each video album in the set of video albums and the user information of the cold-start user, the method further includes:
and determining the users with the number of the watched videos smaller than the preset number threshold value in the preset time length as cold start users. The embodiment of the invention also discloses a video recommendation device, which comprises:
the information acquisition module is used for acquiring the watching information of each video album in the video album set and the user information of the cold start user;
the clustering module is used for clustering the video albums in the video album set according to the watching information of each video album to obtain a plurality of categories of video albums;
the selection module is used for selecting the video albums to be recommended in each category aiming at the video albums in each category;
the matching module is used for calculating the matching degree between the video albums to be recommended and the cold-start user according to the user information and the watching information of each video album to be recommended;
and the recommending module is used for determining a target video album recommended to the cold-start user according to the matching degree.
The embodiment of the invention also discloses an electronic device, which is characterized by comprising: a processor, a memory and a computer program stored on the memory and being executable on the processor, the computer program, when executed by the processor, implementing the steps of the video recommendation method as described above.
The embodiment of the invention also discloses a computer readable storage medium, wherein a computer program is stored on the computer readable storage medium, and when being executed by a processor, the computer program realizes the steps of the video recommendation method.
The embodiment of the invention has the following advantages:
in the embodiment of the invention, video albums of a plurality of categories are obtained by clustering the video albums in the video album set, and video recommendation can be carried out on the video albums to be recommended of each category obtained by clustering; the target video album recommended to the cold-start user is determined by combining the user information of the cold-start user and the watching information of the video album to be recommended, personalized recommendation can be performed for the cold-start user, and therefore the target video album which meets the requirements of the cold-start user better is recommended to the cold-start user in a personalized mode.
Drawings
Fig. 1 is a flowchart illustrating steps of a video recommendation method according to an embodiment of the present invention;
FIG. 2 is a flowchart illustrating steps of another video recommendation method according to an embodiment of the present invention;
2A-2C are flowcharts illustrating sub-steps of another video recommendation method provided by embodiments of the invention;
FIG. 3 is a schematic diagram of training a classification model to be trained according to an embodiment of the present invention;
FIG. 4 is a diagram illustrating an application of a pre-trained classification model according to an embodiment of the present invention;
fig. 5 is a block diagram of a video recommendation apparatus according to an embodiment of the present invention.
Detailed Description
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
The video album can be recommended to the user needing cold start based on the EE algorithm, the method generally needs online calculation to update the recommendation model in real time, however, the online calculation has higher requirements on the performance of the model, and when the sample data is too much, the convergence speed of the model is slower, so that the update speed of the model is reduced, the operating efficiency of the server is influenced, and the quick and effective recommendation for the user is difficult.
One of the core ideas of the embodiment of the invention is that in the embodiment of the invention, video albums of multiple categories are obtained by clustering the video albums in a video album set, and video recommendation can be carried out on the video albums to be recommended of each category obtained by clustering, and the clustering process of the embodiment of the invention does not need to utilize a model, so that online calculation is not needed, and even if the data to be clustered is large, the clustering process does not influence the operating efficiency of a server, so that the video albums are quickly and effectively recommended to cold-start users; the target video album recommended to the cold-start user is determined by combining the user information of the cold-start user and the watching information of the video album to be recommended, and personalized recommendation can be performed on the cold-start user, so that the target video album more meeting the requirements of the cold-start user is personalized recommended to the cold-start user.
Referring to fig. 1, a flowchart illustrating steps of a video recommendation method according to an embodiment of the present invention is shown, where the method specifically includes the following steps:
The video recommendation method provided by the embodiment of the invention can be applied to a recommendation server, and the recommendation server can determine the target video albums recommended to the cold-start users based on the watching information of all the video albums in the video album set and the user information of the cold-start users. When the cold-start user runs the video client at the preset terminal, the target video album recommended by the recommendation server for the cold-start user can be checked.
All videos within a station may be aggregated into a set of video albums, which may include multiple video albums, each of which may include multiple videos. The viewing information of each video album may include multidimensional information such as the total number of clicks, the total number of views, the number of videos, and the total viewing duration of all videos in the video album.
The cold start user may be a video client user that needs a cold start. The user information can be basic user information with static attributes, and can include multidimensional information such as the identity, age, occupation, hobby, and the city where the user is located. In the embodiment of the invention, the multi-dimensional user information of the cold-start user is obtained to recommend the video, so that the target video album which better accords with the characteristics of the user can be recommended to the cold-start user, and the recommendation accuracy rate of the cold-start user is improved.
And 102, clustering the video albums in the video album set according to the watching information of each video album to obtain a plurality of categories of video albums.
The video albums in the video album set can be clustered according to the viewing information of the video albums, so that the video albums are divided to obtain video albums of multiple categories. Where each category of video album may include multiple videos of the same category.
It should be noted that, when a video album is input into a pre-trained video classification model to classify the video album, since the video classification model is trained based on a preset classification rule, the classification result of a part of the video album may be inaccurate, so that the video album recommended to a cold-start user does not fit the user characteristics of the cold-start user when video recommendation is performed based on the classification result of the video album. In the embodiment of the invention, the video albums are clustered, the obtained video album sets of multiple categories are not restricted by preset classification rules, videos to be recommended are selected from the video album sets to be matched with the cold-start user, and the video albums which are attached with the user characteristics of the cold-start user can be recommended to the cold-start user.
In the embodiment of the invention, a preset number of video albums to be recommended can be selected from the plurality of video albums of each category. Illustratively, m video albums to be recommended may be selected from the video albums in the category a, and m video albums to be recommended may be selected from the video albums in the category B.
And 104, calculating the matching degree between the video albums to be recommended and the cold-start user according to the user information and the watching information of each video album to be recommended.
In the embodiment of the invention, the video albums in the video album set are clustered, videos to be recommended are selected from the video albums of multiple categories obtained after clustering, and the matching degree between the video albums to be recommended and the cold-start user is calculated by combining the user information of the cold-start user and the watching information of the video albums to be recommended, so that the operation efficiency of the recommendation server is improved, and the video recommendation is favorably and quickly carried out.
It should be noted that, in the conventional policy recommendation, a candidate group matching with a user requiring cold start is generally defined, and a video album corresponding to the candidate group is recommended to the user requiring cold start. However, in this method, corresponding recommendation strategies need to be formulated for various crowds, and since the user characteristics of the user needing cold start are few, a large amount of debugging needs to be performed on the recommendation strategies corresponding to the various crowds, and video recommendation cannot be performed quickly; moreover, by recommending the video albums corresponding to the matched candidate crowd, the personalized features of the user are easily ignored. In the embodiment of the invention, the recommended video album is determined by calculating the matching degree between the video album to be recommended and the cold-start user, the recommendation strategy is not required to be debugged, and the video recommendation can be quickly carried out; the matching degree is calculated based on the user information of the cold-start user, and personalized recommendation can be performed for different cold-start users.
And 105, determining a target video album recommended to the cold-start user according to the matching degree.
Specifically, after the matching degree between the video to be recommended and the cold-start user is calculated, the video to be recommended with high matching degree with the cold-start user can be determined as a target video album and recommended to the cold-start user. For a plurality of cold start users, based on different user information of different cold start users, the determined target video albums with high matching degree with different cold start users are different, so that the video albums matched with the cold start users can be recommended to the cold start users, and personalized recommendation is performed for the cold start users.
In the embodiment of the invention, video albums of a plurality of categories are obtained by clustering the video albums in the video album set, and video recommendation can be carried out on the video albums to be recommended of each category obtained by clustering; the target video album recommended to the cold-start user is determined by combining the user information of the cold-start user and the watching information of the video album to be recommended, and personalized recommendation can be performed on the cold-start user, so that the target video album more meeting the requirements of the cold-start user is personalized recommended to the cold-start user.
Referring to fig. 2, a flowchart illustrating steps of another video recommendation method according to an embodiment of the present invention is shown, where the method specifically includes the following steps:
In an optional embodiment, before the obtaining the viewing information of each video album in the set of video albums and the user information of the cold-start user, the method further includes: and determining the users with the video watching quantity smaller than the preset quantity threshold value in the preset time length as cold start users.
The recommendation server can identify the users needing cold start, and determines the users with the video watching quantity smaller than a preset quantity threshold value in a preset time length as cold start users. Illustratively, a user who has viewed a number of videos less than or equal to 2 in the last 15 days may be determined as a cold start user. The viewing information of each video album in the set of video albums may be obtained after the cold-start user is determined, and the user information of the cold-start user may be obtained from a database for storing the user information.
And 202, clustering the video albums in the video album set according to the viewing information of each video album to obtain a plurality of categories of video albums.
In an alternative embodiment, referring to fig. 2A, a flow chart of sub-steps of another video recommendation method provided by an embodiment of the present invention is shown, and the step 202 may include the following sub-steps S11 to S12:
and a substep S11, performing feature extraction on the viewing information of each video album to obtain the viewing features of each video album.
The viewing characteristic of the video album may be a video characteristic containing viewing information of the video album. May adopt f doc Representing the viewing characteristics of the video album. For example, if the video album set includes m video albums, after feature extraction is performed on the viewing information of the m video albums, the respective viewing features f of the m video albums can be obtained doc1 ，f doc2 ，…，f docm 。
The number of features N of the viewing feature may be determined by the dimension of the viewing information. Illustratively, if the viewing information of the video album doc1 includes the total number of clicks, the total number of views, the number of videos, and the total viewing duration of all videos in the video album doc1, the feature number N of the viewing feature of the video album doc1 may be 4, for example, f doc1 = [ total number of clicks, total number of views, number of videos, total view time length]＝[1000，22，223，22222]。
And a substep S12, clustering each video album according to the viewing characteristics of each video album to obtain a plurality of categories of video albums.
Specifically, a feature matrix of the video album set can be constructed according to the viewing features of each video album in the video album set, and the feature matrix is clustered to obtain video albums of multiple categories.
Feature extraction is respectively carried out on the watching information of each video album, after the watching features of each video album are obtained, a feature matrix F containing the watching features of each video album can be constructed doc ＝[f doc1 ，f doc2 ，…，f docm ]∈] m×N ，] m×N Can represent F doc Is a feature matrix of m × N, where m is the number of video albums and N is the number of features. The method comprises the steps of determining a characteristic point of each video album in a clustering space based on the viewing characteristic of each video album in a characteristic matrix, adopting a clustering algorithm, randomly selecting C initial clustering centers from a video album set, calculating Euclidean distances between the characteristic point of each video album and each clustering center, finding out the clustering center closest to the characteristic point of the video album, distributing the characteristic point of the video album to the clustering cluster corresponding to the clustering center, calculating an average value of the characteristic points of the video album in each clustering cluster as a new clustering center, and performing next iteration until the clustering center is not changed or the maximum iteration times are reached, so that clustering processing on a plurality of video albums is completed, and video albums of C categories are obtained. At this time, C clustering centers finally corresponding to the video albums of the C categories may be obtained, where the clustering centers may be special samples in the clustering analysis and may be used to represent the categories of the video albums.
At step 204, the distance between each cluster center and each video album is calculated.
After the clustering processing of the plurality of video albums is completed by adopting a clustering algorithm, the clustering center corresponding to the video album of each category does not change any more, at this time, the spatial coordinates of the feature points of each video album and the spatial coordinates of each clustering center can be obtained in a clustering space, and the Euclidean distance between each video album and each clustering center is calculated according to the spatial coordinates of the feature points of each video album and the spatial coordinates of each clustering center. Illustratively, euclidean distances D between doc1 and C cluster centers may be calculated 1,doc1 ，D 2,doc1 ，…，D i,doc1 ，…，D C,doc1 。
And step 205, selecting a plurality of video albums to be recommended in each category according to a preset selection number and the sequence of the distances between each cluster center and each video album from near to far for each cluster center.
For each cluster center, a plurality of to-be-recommended video albums of each category can be selected from the video albums of each category according to the distance between each cluster center and each video album, wherein the same number of to-be-recommended video albums can be selected from the video albums of each category, and when the cluster categories are C, and the number of the to-be-recommended video albums selected from each category is t, the C x t to-be-recommended video albums can be selected. For example, if five video albums doc1, doc2, doc3, doc4, and doc5 in the video album set are clustered to obtain two clustering centers C1 and C2, for the clustering center C1, the distance D between each video album and the clustering center C1 is 1,doc1 ＜D 1,doc2 ＜D 1,doc3 ＜D 1,doc4 ＜D 1,doc5 Selecting two video albums doc1 and doc2 to be recommended of the category a from near to far; for the cluster center C2, the distance D between each video album and the cluster center C2 2,doc4 ＜D 2,doc5 ＜D 2,doc1 ＜D 2,doc2 ＜D 2,doc3 Two video albums to be recommended doc4 and doc5 of category b can be selected from near to far.
Clustering each video album based on the watching characteristics of each video album to obtain the video albums of each category, and determining the category characteristics f of the video albums to be recommended after selecting the video albums to be recommended of each category from the video albums i,doc The belonging category feature may be a feature for indicating a belonging category of the video album. For example, the video feature of the video album doc1 to be recommended may be f 1,doc1 The video characteristic of the video album doc2 to be recommended may be f 1,doc2 The video characteristic of the video album doc4 to be recommended may be f 2,doc4 The video characteristic of the video album doc5 to be recommended may be f 2,doc5 。
And step 206, calculating the matching degree between the video albums to be recommended and the cold-start user according to the user information and the watching information of each video album to be recommended.
In an alternative embodiment, referring to fig. 2B, a flow chart of sub-steps of another video recommendation method provided by the embodiment of the present invention is shown, and the step 206 may include the following sub-steps S21-S24:
and a substep S21, calculating the weight of the category of each video album to be recommended according to the distance between each video album to be recommended and each clustering center.
Each cluster center can correspond to different cluster categories, and the closer the distance between the video album and the cluster center is, the higher the weight of the category to which the video album belongs is. Illustratively, if three cluster centers C1, C2, and C3 are obtained, the cluster type corresponding to the cluster center C1 is a type, the cluster type corresponding to the cluster center C2 is b type, the cluster type corresponding to the cluster center C3 is C type, and distances between the video album doc1 and the cluster centers C1, C2, and C3 are D, respectively 1,doc1 、D 2,doc1 、D 3,doc1 If D is 1,doc1 ＞D 2,doc1 ＞D 3,doc1 Then, it means that the weight of the category c to which the video album doc1 belongs is higher than the weight of the category b to which the video album doc1 belongs, and the weight of the category b to which the video album doc1 belongs is higher than the weight of the category a to which the video album doc1 belongs.
In an alternative embodiment, the substep S21 may comprise: and carrying out normalization calculation on the reciprocal of the distance between each video album to be recommended and each clustering center to obtain the weight of the category to which the video album to be recommended belongs.
Specifically, the reciprocal of the distance between each to-be-recommended video album and each clustering center can be calculated to obtain the preset weight of each category corresponding to each to-be-recommended video album, and the preset weights are normalized to obtain the weight [ w ] of each category corresponding to each to-be-recommended video album 1 ,w 2 ,…,w i ,…,w c ]∈] C ，] C Can represent a C-dimensional feature vector, with weights for each class
Illustratively, the video album doc1 to be recommended and the cluster centers C1, C2,Distance D between C3 1,doc1 、D 2,doc1 、D 3,doc1 Obtaining the preset weight W of doc1 corresponding to class a 1 ＝1/D 1,doc1 Corresponding to the preset weight W of class b 2 ＝1/D 2,doc1 Corresponding to the preset weight W of class c 3 ＝1/D 3,doc1 For each preset weight W 1 、W 2 、W 3 Carrying out normalization calculation to obtain the weight w of the video album doc1 corresponding to the class a 1 Weight w corresponding to class b 2 And weight w corresponding to class c 3 Wherein->
For example, when W 1 ＝1、W 2 ＝10、W 3 If =25, for W 1 、W 2 、W 3 Carrying out normalization calculation to obtain w 1 、w 2 And w 3 ，w 1 ＝1×1/(1+10+25)，w 2 ＝10×1/(1+10+25)，w 3 ＝25×1/(1+10+25)。
In an example, if the video album to be recommended selected according to the order of the distances from near to far is doc1 for the clustering center C3, the video feature of the video album to be recommended doc1 may be f 3,doc1 Match doc1 to the weight w of class c 3 As the weight of the category to which the video album to be recommended belongs.
In one example, the category to which the same video album to be recommended belongs may include a plurality of categories. If the video album to be recommended selected according to the sequence from near to far is doc1 for the clustering center C3, the video feature of the video album to be recommended is f 3,doc1 Let doc1 correspond to the weight w of class c 3 As f 3,doc1 The weight of the category to which it belongs; meanwhile, if doc1 is also the video album to be recommended determined when the clustering center C2 is selected according to the sequence of the distances from near to far, the video characteristic of the video album to be recommended is f 2,doc1 Match doc1 to the weight w of class b 2 As f 2,doc1 The weight of the category to which it belongs.
Step S22, extracting the characteristics of the user information of the cold start user to obtain the user static characteristics of the cold start user; the user static characteristics comprise a VIP identity characteristic, an age characteristic, an occupation characteristic, a preference characteristic and a city characteristic.
The user static feature of the cold-start user may be a user feature containing user information of the cold-start user, and F may be adopted static A user static feature representing a cold start user. The number of features S of the user 'S static features may be determined by the dimension of the user' S information. Illustratively, the user static characteristics F of the cold-start user can be obtained by extracting the characteristics of the user information of the cold-start user static ＝[is vip ,age,occupation,location,hobby,…]∈] S ，] S Can represent an S-dimensional feature vector, is vip The feature of the identity of the guest, that is, whether the user is a VIP (Very Important Person) identity, age may represent an age feature, occupational may represent an occupation feature, location may represent a city feature, and hobby may represent a favorite feature, may be represented.
And a substep S23, inputting the user static characteristics of the cold start user into a pre-trained classification model for processing to obtain the video preference category of the cold start user.
The classification model to be trained can be trained to obtain a pre-trained classification model, and then the user static characteristics of the cold start user are input into the pre-trained classification model to be processed to obtain the video preference category of the cold start user.
In an alternative embodiment, the classification model may be trained by: acquiring training data, wherein the training data comprises user information of a sample user and video preference categories labeled by the sample user; performing feature extraction on the user information of the sample user to obtain user static features of the sample user; and training a classification model by adopting the user static characteristics of the sample user and the video preference categories marked by the sample user to obtain the classification model for identifying the video preference categories.
The classification model may include a fully-connected layer, which may employ DNN (Deep Neural Networks). The static characteristics of the user of the sample user can be input into the classification model to be trained, the output layer of the full connection layer can generate an output result through a ] elu activation function, the output result is input into the softmax layer, the prediction result of the classification model to be trained is obtained, the prediction result can comprise T video preference categories of the sample user, and the classification model can be trained based on the prediction result and the video preference categories marked by the sample user.
Referring to fig. 3, which is a schematic diagram illustrating training of a classification model to be trained according to an embodiment of the present invention, a user static feature F of a sample user is shown static Inputting into a classification model to be trained based on F static For the linear parameter W 1 Training is carried out, a prediction result y' of the classification model to be trained is output through the output layer O, classification loss can be calculated by adopting a loss function based on the predicted video preference category and the labeled video preference category, the loss is reversely propagated to the linear parameters in the network to update the linear parameters, and the process is repeatedly carried out until the classification loss is smaller than an empirical value, wherein the empirical value can be 0.002 exemplarily. Referring to fig. 4, a schematic diagram of applying a pre-trained classification model according to an embodiment of the present invention is shown, in which a user static feature F of a user requiring cold start is used s ` tatic Inputting into a pre-trained classification model based on a linear parameter W 1 Deriving user characteristics F comprising video preference categories of cold-start users user 。
Specifically, the user static feature F of the sample user can be set static Inputting a DNN network to train the classification model to be trained, and obtaining a prediction result y' = softmax (relu (F) of the classification model to be trained static W 1 )O)∈] T Wherein W is 1 ∈] S*C May be referred to as a trainable linearity parameter which,] S*C can represent a matrix containing the matching degree relationship between the S-dimensional user static characteristics and the C cluster categories, and belongs to the field of the user static characteristics] C*T It may be referred to as an output layer,] C*T a matrix containing t pieces of information on video albums to be recommended respectively corresponding to the C cluster categories may be represented,] T can express the prediction result as t-dimensional feature vector and lossThe loss function may be a cross entropy function
y may refer to the true class of the predicted sample.
After the training of the classification model to be trained is finished and the user needing cold start is identified, the user static characteristics of the user needing cold start can be input into the classification model trained in advance, classification is carried out based on the user static characteristics, and the user characteristics f including the video favorite category of the cold start user and output by the output layer can be obtained user ＝relu(F static W 1 )∈] C ，] C The user features may be represented as a C-dimensional feature vector.
And a substep S24, calculating the matching degree between the video album to be recommended and the cold-start user based on the weight of the video favorite category of the cold-start user and the category of the video album to be recommended.
Based on the user characteristics including the video favorite categories of the cold-start users and the weights of the categories of the video albums to be recommended, which are output by the output layer of the pre-trained classification model, the matching degree between the video albums to be recommended and the cold-start users can be calculated.
In the embodiment of the invention, the matching degree between the video album to be recommended and the cold-start user is calculated by combining the user static characteristics of the cold-start user and the watching characteristics of the video album to be recommended, so that the target video album with high matching degree with the cold-start user can be determined, and the target video album conforming to the user static characteristics of the cold-start user can be accurately recommended to the cold-start user.
In an alternative embodiment, referring to fig. 2C, a flow chart of sub-steps of another video recommendation method provided by the embodiment of the present invention is shown, and the sub-step S24 may include the following sub-steps S241-S242:
and a substep S241 of calculating the similarity of the video favorite category of the cold-start user and the category of the video album to be recommended by adopting a similarity calculation method to obtain the similarity between the video album to be recommended and the cold-start user.
Specifically, a cos similarity algorithm may be used,
namely, it is
User characteristics f for video preference categories including cold start users user Video characteristics f of each video album to be recommended i,doc And performing similarity calculation so as to calculate the similarity between each video album to be recommended and the cold start.
And a substep S242, performing weighted average calculation on the similarity by adopting the weight of the category to which the video album to be recommended belongs to obtain the matching degree between the video album to be recommended and the cold-start user.
Wherein, can adopt
And carrying out weighted average calculation on the similarity between the video album to be recommended and the cold-start user to obtain the matching degree between the video album to be recommended and the cold-start user.
For example, in the selected C × t video albums to be recommended, if the video feature of the video album doc1 to be recommended is f 3doc1 The weight of the category is w 3 Then degree of match between doc1 and cold start user
In another example, if the video features of the video doc1 to be recommended include f 3,doc1 And f 2,doc1 The weights are w 3 And w 2 Then degree of match between doc1 and cold start user
In the embodiment of the invention, the matching degree between the video album to be recommended and the cold-start user is calculated based on the weight of the category to which the video album to be recommended belongs and the similarity of the video album to be recommended, and the problems of whether the video album to be recommended and the cold-start user are matched and whether the category to which the video album to be recommended belongs is accurate are considered, so that the target video album which accords with the user static characteristics of the cold-start user is effectively recommended to the cold-start user.
And step 207, determining a target video album recommended to the cold-start user according to the matching degree.
After the matching degree between the videos to be recommended and the cold-start user is calculated, K videos to be recommended can be selected as target video albums and recommended to the cold-start user according to the sequence from high matching degree to low matching degree.
In the embodiment of the invention, video albums of a plurality of categories are obtained by clustering the video albums in the video album set, and video recommendation can be carried out on the video albums to be recommended of each category obtained by clustering; the target video album recommended to the cold-start user is determined by combining the user information of the cold-start user and the watching information of the video album to be recommended, personalized recommendation can be performed for the cold-start user, and therefore the target video album which meets the requirements of the cold-start user better is recommended to the cold-start user in a personalized mode.
In the embodiment of the invention, the matching degree between the video album to be recommended and the cold-start user is calculated by combining the user static characteristics of the cold-start user and the watching characteristics of the video album to be recommended, so that the target video album with high matching degree with the cold-start user can be determined, and the target video album conforming to the user static characteristics of the cold-start user can be accurately recommended to the cold-start user.
In the embodiment of the invention, the matching degree between the video album to be recommended and the cold-start user is calculated based on the weight of the category to which the video album to be recommended belongs and the similarity of the video album to be recommended, and the problems of whether the video album to be recommended and the cold-start user are matched and whether the category to which the video album to be recommended belongs is accurate are considered, so that the target video album which accords with the user static characteristics of the cold-start user is effectively recommended to the cold-start user.
It should be noted that, for simplicity of description, the method embodiments are described as a series of acts or combination of acts, but those skilled in the art will recognize that the present invention is not limited by the illustrated order of acts, as some steps may occur in other orders or concurrently in accordance with the embodiments of the present invention. Further, those skilled in the art will appreciate that the embodiments described in the specification are presently preferred and that no particular act is required to implement the invention.
Referring to fig. 5, a block diagram of a video recommendation apparatus according to an embodiment of the present invention is shown, which may specifically include the following modules:
an information obtaining module 501, configured to obtain viewing information of each video album in the video album set and user information of a cold start user;
a clustering module 502, configured to perform clustering on the video albums in the video album set according to the viewing information of each video album, so as to obtain video albums of multiple categories;
a selecting module 503, configured to select, for each category of video albums, a video album to be recommended in the category;
the matching module 504 is configured to calculate a matching degree between the video album to be recommended and the cold start user according to the user information and the viewing information of each video album to be recommended;
and a recommending module 505, configured to determine, according to the matching degree, a target video album recommended to the cold-start user.
In an optional embodiment, the selecting module includes:
the clustering center acquisition submodule is used for acquiring a clustering center corresponding to the video album set of each category;
the distance calculation submodule is used for calculating the distance between each clustering center and each video album;
and the to-be-recommended album selecting submodule is used for selecting a plurality of to-be-recommended video albums of each category according to the preset selection number and the sequence of the distances between each clustering center and each video album from near to far aiming at each clustering center.
In an alternative embodiment, the matching module includes:
the weight calculation submodule is used for calculating the weight of the category to which each video album to be recommended belongs according to the distance between each video album to be recommended and each clustering center;
the user characteristic extraction submodule is used for extracting the characteristics of the user information of the cold start user to obtain the user static characteristics of the cold start user; the user static characteristics comprise a visitant identity characteristic, an age characteristic, an occupation characteristic, a favorite characteristic and a city characteristic;
the static characteristic classification submodule is used for inputting the user static characteristics of the cold start user into a classification model trained in advance for processing to obtain the video preference category of the cold start user;
and the matching degree operator module is used for calculating the matching degree between the video album to be recommended and the cold-start user based on the weight of the video favorite category of the cold-start user and the category of the video album to be recommended.
In an alternative embodiment, the classification model is trained by:
the training data acquisition module is used for acquiring training data, wherein the training data comprises user information of a sample user and video preference categories labeled by the sample user;
the characteristic extraction module is used for extracting the characteristics of the user information of the sample user to obtain the user static characteristics of the sample user;
and the model training module is used for training the classification model by adopting the user static characteristics of the sample user and the video preference categories marked by the sample user to obtain the classification model for identifying the video preference categories.
In an alternative embodiment, the weight calculation submodule includes:
and the normalization calculation unit is used for performing normalization calculation on the reciprocal of the distance between each video album to be recommended and each clustering center to obtain the weight of the category to which the video album to be recommended belongs.
In an alternative embodiment, the matchmeter operator module comprises:
the similarity calculation unit is used for calculating the similarity of the video favorite category of the cold start user and the category of the video album to be recommended by adopting a similarity calculation method to obtain the similarity between the video album to be recommended and the cold start user;
and the matching degree calculating unit is used for performing weighted average calculation on the similarity by adopting the weight of the category to which the video album to be recommended belongs to obtain the matching degree between the video album to be recommended and the cold-start user.
In an optional embodiment, the cluster processing module includes:
the viewing characteristic extraction submodule is used for extracting the characteristics of the viewing information of each video album to obtain the viewing characteristics of each video album;
and the clustering processing submodule is used for clustering the video albums according to the watching characteristics of the video albums to obtain a plurality of categories of video albums.
In an optional embodiment, before the acquiring the viewing information of each video album in the set of video albums and the user information of the cold-start user, the method further includes:
and the cold start user determining module is used for determining the users watching videos with the number smaller than a preset number threshold value in a preset time length as cold start users.
In the embodiment of the invention, video albums of a plurality of categories are obtained by clustering the video albums in the video album set, and video recommendation can be carried out on the video albums to be recommended of each category obtained by clustering; the target video album recommended to the cold-start user is determined by combining the user information of the cold-start user and the watching information of the video album to be recommended, personalized recommendation can be performed for the cold-start user, and therefore the target video album which meets the requirements of the cold-start user better is recommended to the cold-start user in a personalized mode.
For the device embodiment, since it is basically similar to the method embodiment, the description is simple, and for the relevant points, refer to the partial description of the method embodiment.
An embodiment of the present invention further provides an electronic device, including:
the video recommendation method comprises a processor, a memory and a computer program which is stored in the memory and can run on the processor, wherein when the computer program is executed by the processor, each process of the video recommendation method embodiment is realized, the same technical effect can be achieved, and in order to avoid repetition, the description is omitted here.
The embodiment of the present invention further provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when being executed by a processor, the computer program implements each process of the video recommendation method embodiment, and can achieve the same technical effect, and in order to avoid repetition, the details are not repeated here.
The embodiments in the present specification are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other.
As will be appreciated by one of skill in the art, embodiments of the present invention may be provided as a method, apparatus, or computer program product. Accordingly, embodiments of the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, embodiments of the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
Embodiments of the present invention are described with reference to flowchart illustrations and/or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing terminal to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing terminal to cause a series of operational steps to be performed on the computer or other programmable terminal to produce a computer implemented process such that the instructions which execute on the computer or other programmable terminal provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
While preferred embodiments of the present invention have been described, additional variations and modifications of these embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all such alterations and modifications as fall within the scope of the embodiments of the invention.
Finally, it should also be noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or terminal. Without further limitation, an element defined by the phrases "comprising one of 8230; \8230;" 8230; "does not exclude the presence of additional like elements in a process, method, article, or terminal device that comprises the element.
The video recommendation method, apparatus, electronic device and storage medium provided by the present invention are described in detail above, and a specific example is applied in the text to explain the principle and the implementation of the present invention, and the description of the above embodiment is only used to help understanding the method and the core idea of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed, and in summary, the content of the present specification should not be construed as a limitation to the present invention.
Claims (11)
1. A method for video recommendation, comprising:
acquiring the watching information of each video album in the video album set and the user information of a cold start user;
clustering the video albums in the video album set according to the watching information of each video album to obtain a plurality of categories of video albums;
selecting a video album to be recommended in each category aiming at the video albums in each category;
calculating the matching degree between the video album to be recommended and the cold start user according to the user information and the watching information of each video album to be recommended;
and determining a target video album recommended to the cold-start user according to the matching degree.
2. The method as claimed in claim 1, wherein the selecting a video album to be recommended for each category from the video albums of the category comprises:
acquiring a clustering center corresponding to each category of video albums;
calculating the distance between each cluster center and each video album;
and aiming at each cluster center, selecting a plurality of video albums to be recommended in each category according to a preset selection number according to the sequence of the distance between each cluster center and each video album from near to far.
3. The method as claimed in claim 2, wherein the calculating the matching degree between the video album to be recommended and the cold-start user according to the user information and the viewing information of each video album to be recommended comprises:
calculating the weight of the category of each video album to be recommended according to the distance between each video album to be recommended and each clustering center;
performing feature extraction on the user information of the cold start user to obtain the user static feature of the cold start user; the user static characteristics comprise a VIP identity characteristic, an age characteristic, an occupation characteristic, a preference characteristic and a city characteristic;
inputting the user static characteristics of the cold start user into a pre-trained classification model for processing to obtain the video preference category of the cold start user;
and calculating the matching degree between the video album to be recommended and the cold-start user based on the weight of the video favorite category of the cold-start user and the category of the video album to be recommended.
4. The method of claim 3, wherein the classification model is trained by:
acquiring training data, wherein the training data comprises user information of a sample user and video preference categories labeled by the sample user;
performing feature extraction on the user information of the sample user to obtain user static features of the sample user;
and training a classification model by adopting the user static characteristics of the sample user and the video preference categories marked by the sample user to obtain the classification model for identifying the video preference categories.
5. The method as claimed in claim 3, wherein the calculating the weight of the category to which each video album to be recommended belongs according to the distance between each video album to be recommended and the respective cluster center comprises:
and carrying out normalization calculation on the reciprocal of the distance between each video album to be recommended and each clustering center to obtain the weight of the category to which the video album to be recommended belongs.
6. The method as claimed in claim 5, wherein the calculating the matching degree between each video album to be recommended and the cold-start user based on the weight of the video preference category of the cold-start user and the category to which the video album to be recommended belongs comprises:
performing similarity calculation on the video favorite category of the cold start user and the category of the video album to be recommended by adopting a similarity calculation method to obtain the similarity between the video album to be recommended and the cold start user;
and performing weighted average calculation on the similarity by adopting the weight of the category to which the video album to be recommended belongs to obtain the matching degree between the video album to be recommended and the cold-start user.
7. The method as claimed in claim 1, wherein the clustering the video albums in the video album set according to the viewing information of the video albums to obtain video albums of multiple categories comprises:
performing feature extraction on the viewing information of each video album to obtain the viewing features of each video album;
and clustering the video albums according to the watching characteristics of the video albums to obtain video albums of multiple categories.
8. The method of claim 1, further comprising, prior to the obtaining viewing information for each video album in the set of video albums and user information for a cold-start user:
and determining the users with the video watching quantity smaller than the preset quantity threshold value in the preset time length as cold start users.
9. A video recommendation apparatus, characterized in that the apparatus comprises:
the information acquisition module is used for acquiring the watching information of each video album in the video album set and the user information of the cold start user;
the clustering module is used for clustering the video albums in the video album set according to the watching information of each video album to obtain a plurality of categories of video albums;
the selection module is used for selecting the video albums to be recommended in each category aiming at the video albums in each category;
the matching module is used for calculating the matching degree between the video albums to be recommended and the cold-start user according to the user information and the watching information of each video album to be recommended;
and the recommending module is used for determining a target video album recommended to the cold-start user according to the matching degree.
10. An electronic device, comprising: processor, memory and computer program stored on the memory and executable on the processor, the computer program when executed by the processor implementing the steps of the video recommendation method of any one of claims 1-8.
11. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the video recommendation method according to any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211599774.1A CN115878897A (en) | 2022-12-14 | 2022-12-14 | Video recommendation method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211599774.1A CN115878897A (en) | 2022-12-14 | 2022-12-14 | Video recommendation method and device, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115878897A true CN115878897A (en) | 2023-03-31 |
Family
ID=85767287
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211599774.1A Pending CN115878897A (en) | 2022-12-14 | 2022-12-14 | Video recommendation method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115878897A (en) |
-
2022
- 2022-12-14 CN CN202211599774.1A patent/CN115878897A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110046304B (en) | User recommendation method and device | |
| CN108921221B (en) | User feature generation method, device, equipment and storage medium | |
| CN111797321B (en) | Personalized knowledge recommendation method and system for different scenes | |
| CN106815252B (en) | Searching method and device | |
| CN106651542B (en) | Article recommendation method and device | |
| CN110717099B (en) | Method and terminal for recommending film | |
| CN111382307B (en) | Video recommendation method, system and storage medium based on deep neural network | |
| CN111209476A (en) | Recommendation method, model generation method, device, medium and equipment | |
| CN110008397B (en) | Recommendation model training method and device | |
| CN113505204B (en) | Recall model training method, search recall device and computer equipment | |
| CN111506820B (en) | Recommendation model, recommendation method, recommendation device, recommendation equipment and recommendation storage medium | |
| CN111506821B (en) | Recommendation model, recommendation method, recommendation device, recommendation equipment and recommendation storage medium | |
| CN110688565B (en) | Next item recommendation method based on multidimensional Hox process and attention mechanism | |
| CN107577736B (en) | File recommendation method and system based on BP neural network | |
| CN110222838B (en) | Document sorting method and device, electronic equipment and storage medium | |
| WO2018118314A1 (en) | Mixed data fingerprinting with principal components analysis | |
| CN113449011A (en) | Big data prediction-based information push updating method and big data prediction system | |
| CN113656699B (en) | User feature vector determining method, related equipment and medium | |
| CN113704620A (en) | User label updating method, device, equipment and medium based on artificial intelligence | |
| Guan et al. | Enhanced SVD for collaborative filtering | |
| CN115878897A (en) | Video recommendation method and device, electronic equipment and storage medium | |
| CN112818196B (en) | Data processing method, equipment, electronic device and storage medium based on electronic learning platform | |
| CN111400413B (en) | Method and system for determining category of knowledge points in knowledge base | |
| CN114610913A (en) | Recommendation method and recommendation model training method for multimedia data and related equipment | |
| CN104636489B (en) | The treating method and apparatus of attribute data is described |
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 |