CN111339355A - Video recommendation method and system - Google Patents
Video recommendation method and system Download PDFInfo
- Publication number
- CN111339355A CN111339355A CN202010434612.7A CN202010434612A CN111339355A CN 111339355 A CN111339355 A CN 111339355A CN 202010434612 A CN202010434612 A CN 202010434612A CN 111339355 A CN111339355 A CN 111339355A
- Authority
- CN
- China
- Prior art keywords
- long
- video
- user
- sample
- recommended
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/73—Querying
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/74—Browsing; Visualisation therefor
- G06F16/743—Browsing; Visualisation therefor a collection of video files or sequences
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/957—Browsing optimisation, e.g. caching or content distillation
Abstract
The invention provides a video recommendation method and a video recommendation system, which are used for acquiring long and short video characteristics and user characteristics of a user; inputting the user characteristics and the long and short video characteristics into a preset recommendation model for click rate prediction to obtain the predicted click rate of each long video to be recommended and the predicted click rate of each short video to be recommended; and sequencing all the long videos to be recommended and the short videos to be recommended according to the predicted click rate of each long video to be recommended and the predicted click rate of each short video to be recommended, and feeding back sequencing results to the user. In the scheme, the user characteristics and the long and short video characteristics obtained by combining the long video characteristics and the short video characteristics are input into a pre-trained recommendation model to obtain the predicted click rates of the long video to be recommended and the short video to be recommended. And sequencing the long video to be recommended and the short video to be recommended according to the predicted click rate, and feeding back a sequencing result to the user, so that the recommendation of different types of information is realized, the user experience is improved, and the accuracy of information recommendation is improved.
Description
Technical Field
The invention relates to the technical field of data processing, in particular to a video recommendation method and system.
Background
With the development of the internet, various application software has been developed. During the process of using the application software by the user, the application software provides personalized recommendation service for the user.
At present, the way of providing personalized recommendation service to users is as follows: and filtering the information according to the preference of the user by using a recommendation algorithm according to the historical behavior data of the user to obtain the preferred information of the user. However, the current personalized recommendation service can only recommend a single type of information, and as the information types are continuously increased and the characteristics of each type of information are different, the current personalized recommendation service cannot meet the actual requirements of users and can not accurately recommend multiple types of information, the user experience is poor and the accuracy of information recommendation is poor.
Disclosure of Invention
In view of this, embodiments of the present invention provide a video recommendation method and system to solve the problems of poor user experience, poor accuracy of information recommendation, and the like in the current personalized recommendation service.
In order to achieve the above purpose, the embodiments of the present invention provide the following technical solutions:
the first aspect of the embodiments of the present invention discloses a video recommendation method, including:
acquiring long and short video characteristics and user characteristics of a user, wherein the long and short video characteristics are obtained by combining long video characteristics of a plurality of long videos to be recommended and short video characteristics of a plurality of short videos to be recommended;
inputting the user characteristics and the long and short video characteristics into a preset recommendation model for click rate prediction to obtain the predicted click rate of each long video to be recommended and the predicted click rate of each short video to be recommended, and training a machine learning model according to a training data set obtained in advance by the recommendation model;
and sequencing all the long videos to be recommended and the short videos to be recommended according to the predicted click rate of each long video to be recommended and the predicted click rate of each short video to be recommended, and feeding back sequencing results to the user.
Preferably, the acquiring the long and short video features and the user features of the user includes:
acquiring historical long and short video characteristics and real-time long and short video characteristics, and fusing the historical long and short video characteristics and the real-time long and short video characteristics to obtain long and short video characteristics;
the method comprises the steps of obtaining historical user characteristics and real-time user characteristics corresponding to a user, and fusing the historical user characteristics and the real-time user characteristics to obtain the user characteristics corresponding to the user.
Preferably, the process of acquiring the training data set comprises:
obtaining a sample long video image corresponding to the sample long video and a sample short video portrait corresponding to the sample short video, and obtaining a sample user portrait and a sample context behavior portrait corresponding to the sample user;
performing data preprocessing on the sample long video portrait, the sample short video portrait, the sample user portrait and the sample context behavior portrait to obtain an original characteristic data set;
and carrying out feature engineering processing on the original feature data set to obtain a training data set, wherein the feature engineering processing at least comprises feature extraction processing and feature transformation processing.
Preferably, the process of training the machine learning model to obtain the recommendation model according to the pre-acquired training data set includes:
dividing the training data set into a training set, a verification set and a test set according to a preset division ratio;
and training the machine learning model according to the training set, the verification set and the test set and in combination with the click behavior and the exposure behavior of the sample user on the sample long video and the sample short video until the machine learning model is converged to obtain a recommendation model.
Preferably, after all the long videos to be recommended and the short videos to be recommended are sorted and the sorting result is fed back to the user, the method further includes:
and updating and optimizing the recommendation model according to the clicking behavior and the exposure behavior of the user on the long video to be recommended and the short video to be recommended.
A second aspect of an embodiment of the present invention discloses a video recommendation system, including:
the device comprises an acquisition unit, a recommendation unit and a recommendation unit, wherein the acquisition unit is used for acquiring long and short video characteristics and user characteristics of a user, and the long and short video characteristics are obtained by combining long video characteristics of a plurality of long videos to be recommended and short video characteristics of a plurality of short videos to be recommended;
the processing unit is used for inputting the user characteristics and the long and short video characteristics into a preset recommendation model for click rate prediction to obtain the predicted click rate of each long video to be recommended and the predicted click rate of each short video to be recommended, and the recommendation model is obtained by training a machine learning model according to a training data set acquired in advance;
and the sequencing unit is used for sequencing all the long videos to be recommended and the short videos to be recommended according to the predicted click rate of each long video to be recommended and the predicted click rate of each short video to be recommended, and feeding back sequencing results to the user.
Preferably, the obtaining unit is specifically configured to: the method comprises the steps of obtaining historical long and short video features and real-time long and short video features, fusing the historical long and short video features and the real-time long and short video features to obtain long and short video features, obtaining historical user features and real-time user features corresponding to users, and fusing the historical user features and the real-time user features to obtain user features corresponding to the users.
Preferably, the processing unit for acquiring a training data set comprises:
the acquisition module is used for acquiring a sample long video image corresponding to the sample long video and a sample short video portrait corresponding to the sample short video, and acquiring a sample user portrait and a sample context behavior portrait corresponding to a sample user;
the preprocessing module is used for preprocessing the sample long video portrait, the sample short video portrait, the sample user portrait and the sample context behavior portrait to obtain an original characteristic data set;
and the characteristic engineering processing module is used for carrying out characteristic engineering processing on the original characteristic data set to obtain a training data set, wherein the characteristic engineering processing at least comprises characteristic extraction processing and characteristic conversion processing.
Preferably, the processing unit for training the machine learning model to obtain the recommended model includes:
the dividing module is used for dividing the training data set into a training set, a verification set and a test set according to a preset dividing proportion;
and the training module is used for training the machine learning model according to the training set, the verification set and the test set and combining the click behavior and the exposure behavior of the sample user on the sample long video and the sample short video until the machine learning model is converged to obtain a recommendation model.
Preferably, the system further comprises:
and the updating unit is used for updating and optimizing the recommendation model according to the clicking behavior and the exposure behavior of the user on the long video to be recommended and the short video to be recommended.
Based on the video recommendation method and system provided by the embodiment of the invention, the method comprises the following steps: acquiring long and short video characteristics and user characteristics of a user; inputting the user characteristics and the long and short video characteristics into a preset recommendation model for click rate prediction to obtain the predicted click rate of each long video to be recommended and the predicted click rate of each short video to be recommended; and sequencing all the long videos to be recommended and the short videos to be recommended according to the predicted click rate of each long video to be recommended and the predicted click rate of each short video to be recommended, and feeding back sequencing results to the user. In the scheme, the user characteristics and the long and short video characteristics obtained by combining the long video characteristics and the short video characteristics are input into a pre-trained recommendation model to obtain the predicted click rates of the long video to be recommended and the short video to be recommended. And sequencing the long video to be recommended and the short video to be recommended according to the predicted click rate, and feeding back a sequencing result to the user, so that the recommendation of different types of information is realized, the user experience is improved, and the accuracy of information recommendation is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
Fig. 1 is a flowchart of a video recommendation method according to an embodiment of the present invention;
FIG. 2 is a flowchart of acquiring a training data set according to an embodiment of the present invention;
fig. 3 is another flowchart of a video recommendation method according to an embodiment of the present invention;
fig. 4 is a block diagram of a video recommendation system according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In this application, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
The background art can know that the current mode of providing personalized recommendation service for users can only recommend single type of information, and along with the continuous increase of information types and the different characteristics of each type of information, the current personalized recommendation service can not meet the actual requirements of users and can not accurately recommend multiple types of information, and the problems of poor user experience, poor accuracy of information recommendation and the like exist.
Therefore, the embodiment of the invention provides a video recommendation method and system, wherein a recommendation model is obtained by pre-training, and the user characteristics and the long and short video characteristics obtained by combining the long video characteristics and the short video characteristics are input into the recommendation model to obtain the predicted click rates of the long video to be recommended and the short video to be recommended. And sequencing the long video to be recommended and the short video to be recommended according to the predicted click rate, and feeding back sequencing results to the user to realize the recommendation of different types of information so as to improve the user experience and the accuracy of information recommendation.
Referring to fig. 1, a flowchart of a video recommendation method provided by an embodiment of the present invention is shown, where the video recommendation method includes the following steps:
step S101: and acquiring long and short video characteristics and user characteristics of a user.
It should be noted that, a recommendation request sent by a user is received in advance, and the recommendation request at least includes a user ID corresponding to the user. After receiving the recommendation request, obtaining the user characteristics and the long and short video characteristics corresponding to the user ID from a memory database (e.g., a redis database).
Before receiving a recommendation request, combining long video features of a plurality of long videos to be recommended and short video features of a plurality of short videos to be recommended in advance to obtain long and short video features, and storing the long and short video features in an in-memory database.
It can be understood that, in order to ensure that the obtained long and short video features and the user features are real-time latest and accurate, in the process of obtaining the long and short video features and the user features, the historical long and short video features and the real-time long and short video features are obtained, and the long and short video features are obtained by fusing the historical long and short video features and the real-time long and short video features.
Namely, according to a preset period, the historical long and short video features and the real-time long and short video features are fused to obtain the long and short video features, and the real-time performance and the accuracy of the long and short video features are guaranteed. The long and short video features are stored in a memory database in advance, and are directly read after a recommendation request of a user is received, so that the reading and writing speed is improved.
And acquiring historical user characteristics and real-time user characteristics corresponding to the user, and fusing the historical user characteristics and the real-time user characteristics to obtain the user characteristics corresponding to the user.
That is, the behavior data of the user (e.g., the current preference data of the user) is obtained in real time, and the real-time user characteristics corresponding to the user are generated according to the behavior data obtained in real time. And the historical user characteristics and the real-time user characteristics are fused to obtain the user characteristics corresponding to the user, so that the real-time performance and the accuracy of the user characteristics are ensured. The user characteristics are stored in the memory database in advance, and are directly read after the recommendation request of the user is received, so that the reading and writing speed is improved.
Step S102: and inputting the user characteristics and the long and short video characteristics into a preset recommendation model to predict click rate, so as to obtain the predicted click rate of each long video to be recommended and the predicted click rate of each short video to be recommended.
It should be noted that, a sample long video image corresponding to the sample long video and a sample short video image corresponding to the sample short video are collected in advance, and a sample user portrait and a sample context behavior portrait corresponding to the sample user are collected.
And constructing a training data set for training the machine learning model according to the sample long video portrait, the sample short video portrait, the sample user portrait and the sample context behavior portrait.
It should be further noted that the sample context behavior representation refers to: and the scene of the video clicking behavior of the sample user comprises the specific time when the sample user clicks the video, the video position clicked by the sample user, the last video watched before the sample user clicks the video and the next video watched after the sample user clicks the video.
The sample long video representation and the sample short video representation are generated by processing the long video and the short video in a database (a database storing the full amount of long video and short video).
It is understood that after the training data set is obtained, the training data set is divided into a training set, a verification set and a test set according to a preset division ratio, for example: when the training data set is divided, the training data set is divided according to the time sequence and the division ratio to obtain a training set, a verification set and a test set.
It should be noted that the training set is a data sample for training the machine learning model, the validation set is a sample set for adjusting model parameters during training the machine learning model, and the test set is used for evaluating generalization ability (prediction ability of click rate) of the trained machine learning model. It is understood that the training set, validation set, and test set are subject to the same data distribution.
And generating positive and negative samples according to the training set, the verification set and the test set and in combination with the click behavior and the exposure behavior of the sample user on the sample long video and the sample short video. And training the machine learning model by using the positive and negative samples until the machine learning model converges to obtain a recommendation model.
It should be noted that the types of machine learning models include, but are not limited to: FM linear model, LightGBM tree model, deep FM learning model.
In the process of training the machine learning model, the machine learning model is subjected to super-parameter adjustment and optimization by using Grid Search, Random Search or Hyperopt and the like, and a recommended model is determined by selecting the best super-parameter combination.
In the process of specifically implementing step S102, the user characteristics and the long and short video characteristics are input into the recommendation model, and the click rate of the user on each long video to be recommended and each short video to be recommended (the degree of preference of the user on each long video to be recommended and each short video to be recommended) is predicted by using the recommendation model, so as to obtain the predicted click rate of each long video to be recommended and the predicted click rate of each short video to be recommended.
Step S103: and sequencing all the long videos to be recommended and the short videos to be recommended according to the predicted click rate of each long video to be recommended and the predicted click rate of each short video to be recommended, and feeding back sequencing results to the user.
In the process of specifically implementing the step S103, after the predicted click rate of each long video to be recommended and the predicted click rate of each short video to be recommended are obtained by using the recommendation model, all the long videos to be recommended and the short videos to be recommended are sorted according to the order of the predicted click rates from high to low, and the sorting result is fed back to the user.
It can be understood that after all the long videos to be recommended and the short videos to be recommended are ranked, the top n videos to be recommended (the videos to be recommended are the long videos to be recommended or the short videos to be recommended) in the ranking result are recommended to the user, n is a positive integer, and the specific numerical value is set according to the actual situation.
Preferably, in order to ensure the prediction accuracy of the recommendation model, the recommendation model is updated and optimized (parameters of the recommendation model are adjusted and optimized) according to the click behavior and the exposure behavior of the long video to be recommended and the short video to be recommended of the user.
And performing AB test (A/BTesting) on the recommendation model (the model before updating and optimizing) and the recommendation model after updating and optimizing, and evaluating the recommendation model (the model before updating and optimizing) and the recommendation model after updating and optimizing according to a preset evaluation index.
And if the evaluation effect of the recommendation model before updating and optimizing is better than that of the recommendation model after updating and optimizing, continuing to use the recommendation model before updating and optimizing to predict the click rate in the subsequent use process.
And if the evaluation effect of the updated and optimized recommendation model is better than that of the recommendation model before updating and optimizing, predicting the click rate in the subsequent use process by using the updated and optimized recommendation model.
In the embodiment of the invention, the machine learning model is trained in advance by using the training data set to obtain the recommendation model. And inputting the user characteristics and the long and short video characteristics obtained by combining the long video characteristics and the short video characteristics into a recommendation model to obtain the predicted click rate of the long video to be recommended and the predicted click rate of the short video to be recommended. And sequencing the long video to be recommended and the short video to be recommended according to the predicted click rate, feeding back a sequencing result to the user, recommending different types of information (the long video and the short video), and improving the user experience and the accuracy of information recommendation.
The above-mentioned process for acquiring the training data set in step S102 in fig. 1 according to the embodiment of the present invention is shown in fig. 2, which is a flowchart for acquiring the training data set provided in the embodiment of the present invention, and includes the following steps:
step S201: and obtaining a sample long video image corresponding to the sample long video and a sample short video portrait corresponding to the sample short video, and obtaining a sample user portrait and a sample context behavior portrait corresponding to the sample user.
In the process of implementing step S201 specifically, the long video and the short video in the database (the database storing the full amount of long video and short video) are processed to obtain a sample long video portrait and a sample short video portrait.
And processing the user behavior log corresponding to the sample user by using a big data tool (such as hive and spark) to obtain a sample user image and a sample context behavior image corresponding to the sample user.
Step S202: and carrying out data preprocessing on the sample long video portrait, the sample short video portrait, the sample user portrait and the sample context behavior portrait to obtain an original characteristic data set.
In the process of the specific implementation step S202, data preprocessing is performed on the sample long video portrait, the sample short video portrait, the sample user portrait, and the sample context behavior portrait, respectively, to obtain an original feature data set containing a result of the data preprocessing. That is, the raw feature data set contains: the method comprises the steps of preprocessing a sample long video image data, preprocessing a sample short video image data, preprocessing a sample user image data and preprocessing a sample context behavior image data.
A specific process for data pre-processing a sample long video representation, a sample short video representation, a sample user representation, and a sample contextual behavioral representation is described below.
Preprocessing data of the sample long video image and the sample short video image: unifying the data formats of the sample long video portrait and the sample short video portrait, and filling a fixed field (a preset field) in a field with an empty attribute in the sample long video portrait and the sample short video portrait, wherein the field filled with the fixed field represents an empty field.
Data pre-processing of sample user portrayal: and formatting the clicking behaviors of the sample users in the video categories and other attributes, and filtering the sample users without the clicking behaviors.
It can be understood that, since the click behavior is an original log corresponding to the sample user, the log does not conform to the storage format of the sample user portrait, and therefore, the click behavior needs to be formatted into a format conforming to the preset specification.
For a portion of sample users that may not have click behavior, or that may have click behavior but viewed the video for less than a predefined viewing time (indicating not viewed), this portion of sample users is filtered out.
Preprocessing data of the sample context behavior portrait: and screening the behavior data of the sample user within a preset time period (for example, according to the latest 4 weeks), and generating a sample context behavior portrait corresponding to the sample user by using the screened behavior data.
Step S203: and carrying out characteristic engineering processing on the original characteristic data set to obtain a training data set.
In the process of implementing step S203 specifically, feature engineering processing is performed on the data in the original feature data set to obtain a formatting feature for training the machine learning model, and the generated formatting feature is the training data set.
It should be noted that the feature engineering process at least includes a feature extraction process and a feature transformation process, and in order to better explain how to perform the feature engineering process on the original feature data set, the feature engineering process is exemplified by the following contents.
Filling null values: and filling a mode or default value into the class type characteristic vacancy values in the original characteristic data set, and filling 0 or mean value into the continuity characteristic vacancy values in the original characteristic data set.
Characteristic transformation: the features are generated by discretizing one-hot (one-hot coding) on discrete variables (class type features) and continuous variables in an original feature data set, and the features are generated by normalizing the continuous variables.
Converting the characteristics of the attributes such as titles, labels and the like of long and short videos (the long videos and the short videos after the uniform data format): extracting fusion label characteristics of the long and short videos, Embedding the fusion label characteristics of the long and short videos by using a label word vector model corresponding to the long and short videos obtained through pre-training to obtain label vectors corresponding to the long and short videos, and clustering (or normalizing) the label vectors to generate characteristics corresponding to the long and short videos. Extracting the titles of the long and short videos, segmenting the titles, processing the titles of the long and short videos by using a pre-trained entry word vector model corresponding to the long and short videos to obtain entry word vectors, and clustering (or normalizing) the entry word vectors to generate features corresponding to the long and short videos.
And (3) converting the characteristics of the category clicking behaviors of the sample users in the long and short videos: and fusing the category clicking behaviors of the sample user in the long and short videos to generate the feature of the category behavior clicking weight vector. And clustering the click weight vector characteristics of the category behaviors by using a clustering model obtained by pre-training to generate clustering characteristics.
Feature transformation of long and short video tag behaviors of sample users: and converting the long and short video label behaviors of the sample user into vectors by using a label word vector model obtained by pre-training, and clustering the vectors by using a pre-trained clustering model to generate clustering characteristics.
Performing crowd classification on sample users: and classifying the crowd of the sample user by using the basic information such as the mobile phone model and the geographic position of the sample user. The method comprises the steps of calculating tf-idf (term frequency-inverse cluster frequency) by using an Application (APP) list of a sample user, and carrying out crowd clustering on the sample user through a clustering model (such as a Kmeans clustering model) obtained through pre-training to obtain corresponding clustering characteristics.
Dividing click behaviors of long and short videos: the long video and the short video have different physical durations and different attributes, so that the long video and the short video are divided according to the distribution of the click duration to determine the playing completion degree characteristics of the long video and the short video, and the playing completion degree of the target long video and the target short video are weighted.
Converting the heat characteristics of the long and short videos: since the hits of short videos are usually millions (even hundreds of millions) times, and the hits of long videos are usually tens of thousands of times due to the property limit of long videos, the heat calculation of short videos and long videos is different. And (4) displaying before filtering the short video, exposing the sample user once, and calculating the heat characteristic of the short video according to the click number and the exposure number. The long video can be displayed for multiple times, exposure and de-duplication processing are carried out every day, and the heat characteristic of the long video is calculated by combining the total clicks of the episode under the long video. Similarly, the average playing duration can be obtained by dividing the accumulated playing duration of the long and short videos by the total clicks, so that the playing completion degree of the long and short videos is calculated, that is, the evaluation of the sample user on the quality of the long and short videos is used as the heat characteristic.
It can be understood that after the features are determined in the above manners, in order to avoid redundancy caused by the existence of a large number of features, the features need to be selected, and a specific feature selection manner is as follows: indexes such as correlation coefficients between the respective features and target values (target values are 0 and 1, 0 indicates no click, and 1 indicates click, and the purpose is to screen effective and important features) are calculated, and the respective features are evaluated. And selecting important features according to the evaluation result for verification, determining baseline, and then optimizing the features by using indexes such as model training, correlation coefficients and the like.
It should be noted that, the above mentioned label word vector model and the above mentioned entry word vector model may be obtained by using word2vec model or bert model training, and parameters of the label word vector model and the entry word vector model are adjusted through word vector quality inspection and online verification.
For the application of the clustering model, the distribution of the category number is continuously adjusted by checking the distribution of the category number, so that most samples are prevented from being clustered into one category, and the clustering model and the characteristics are optimized according to the checking result of checking the distribution of the category number.
And constructing a training data set by using the characteristics determined by the mode, and training the machine learning model by using the training data set to obtain a recommendation model.
In the embodiment of the invention, data preprocessing is carried out on a sample long video image, a sample short video image, a sample user image and a sample context behavior image to obtain an original characteristic data set. And performing characteristic engineering processing such as characteristic extraction processing, characteristic conversion processing and the like on the original characteristic data set to obtain a training data set containing formatting characteristics for training a machine learning model. The machine learning model is trained by the training data set to obtain a recommendation model, and personalized recommendation services for different types of information are provided for the user through the recommendation model, so that the user experience is improved, and the accuracy of information recommendation is improved.
To better explain the contents shown in fig. 1 and fig. 2 of the above embodiments of the present invention, fig. 3 is used for illustration, and it should be noted that fig. 3 is used for illustration only.
Referring to fig. 3, another flowchart of a video recommendation method according to an embodiment of the present invention is shown, including the following steps:
step S301: and in an off-line mode, acquiring a training data set, and training the machine learning model by using the training data set to obtain a recommendation model.
In the process of the specific implementation step S301, a sample long video portrait, a sample short video portrait, a sample user portrait and a sample context behavior portrait are obtained, and data preprocessing and feature engineering processing are performed on the sample long video portrait, the sample short video portrait, the sample user portrait and the sample context behavior portrait respectively to obtain a training data set.
Wherein, step S301 includes substeps S3011 to substep S3014.
Substep S3011: and acquiring a sample long video image and a sample short video image from a video data source, and processing a user behavior log of a sample user by using hive to obtain a sample user image and a sample context behavior image.
The video data source stores long video images (also storing long video features) corresponding to a plurality of long videos and short video images (also storing short video features) corresponding to a plurality of short videos.
Substep S3012: and respectively carrying out data preprocessing on the sample long video portrait, the sample short video portrait, the sample user portrait and the sample context behavior portrait to obtain an original characteristic data set.
In the process of implementing the sub-step S3012, the data preprocessing process is as described in step S202 of fig. 2 in the embodiment of the present invention.
Substep S3013: and performing feature engineering processing on the original feature data set to obtain a training data set containing formatting features, and storing the training data set to a Distributed File System (hdfs).
In the process of implementing the sub-step S3013, the process of feature engineering is performed, as described in step S203 in fig. 2 in the embodiment of the present invention.
Substep S3014: and training the machine learning model by using the training data set in the hdfs to obtain a recommendation model and storing the recommendation model offline.
Step S302: and acquiring real-time user characteristics corresponding to the user, acquiring real-time long and short video characteristics from a video data source in the near-line mode, and acquiring historical long and short video characteristics and historical user characteristics from data processed in the off-line mode. And fusing historical long and short video features and real-time long and short video features to obtain long and short video features, and fusing historical user features and real-time user features to obtain user features corresponding to the user.
Step S303: in an online mode, inputting user characteristics and long and short video characteristics into a recommendation model to predict click rate, obtaining the predicted click rate of each long video to be recommended and the predicted click rate of each short video to be recommended, sequencing all the long videos to be recommended and the short videos to be recommended, and feeding sequencing results back to a user.
Step S303 includes sub-steps S3031 to S3034, which are as follows.
Substep S3031: and receiving a recommendation request sent by a user.
Substep S3032: and obtaining user characteristics and long and short video characteristics, loading a recommendation model by using C + +, inputting the user characteristics and the long and short video characteristics into the recommendation model to predict click rate, and obtaining the predicted click rate of each long video to be recommended and the predicted click rate of each short video to be recommended.
Substep S3033: and sequencing all the long videos to be recommended and the short videos to be recommended according to the high and low sequences of the predicted click rate.
Substep S3034: and feeding back the sequencing result to the user.
It should be noted that, the execution principle of each step in fig. 3 can refer to the contents in fig. 1 and fig. 2 in the above embodiment of the present invention, and details are not repeated herein.
Corresponding to the video recommendation method provided by the embodiment of the present invention, referring to fig. 4, an embodiment of the present invention further provides a structural block diagram of a video recommendation system, where the video recommendation system includes: an acquisition unit 401, a processing unit 402, and a sorting unit 403;
the obtaining unit 401 is configured to obtain long and short video features and user features of a user, where the long and short video features are obtained by combining long video features of a plurality of long videos to be recommended and short video features of a plurality of short videos to be recommended.
In a specific implementation, the obtaining unit 401 is specifically configured to: the method comprises the steps of obtaining historical long and short video features and real-time long and short video features, fusing the historical long and short video features and the real-time long and short video features to obtain long and short video features, obtaining historical user features and real-time user features corresponding to users, and fusing the historical user features and the real-time user features to obtain user features corresponding to the users.
The processing unit 402 is configured to input the user characteristics and the long and short video characteristics into a preset recommendation model for click rate prediction, so as to obtain a predicted click rate of each long video to be recommended and a predicted click rate of each short video to be recommended, where the recommendation model is obtained by training a machine learning model according to a training data set acquired in advance.
The sorting unit 403 is configured to sort all the long videos to be recommended and the short videos to be recommended according to the predicted click rate of each long video to be recommended and the predicted click rate of each short video to be recommended, and feed back a sorting result to the user.
In the embodiment of the invention, the machine learning model is trained in advance by using the training data set to obtain the recommendation model. And inputting the user characteristics and the long and short video characteristics obtained by combining the long video characteristics and the short video characteristics into a recommendation model to obtain the predicted click rate of the long video to be recommended and the predicted click rate of the short video to be recommended. And sequencing the long video to be recommended and the short video to be recommended according to the predicted click rate, feeding back a sequencing result to the user, recommending different types of information (the long video and the short video), and improving the user experience and the accuracy of information recommendation.
Preferably, in connection with the content shown in fig. 4, the processing unit 402 for obtaining the training data set comprises: the system comprises an acquisition module, a preprocessing module and a characteristic engineering processing module, wherein the execution principle of each module is as follows:
the acquisition module is used for acquiring a sample long video image corresponding to the sample long video and a sample short video portrait corresponding to the sample short video, and acquiring a sample user portrait and a sample context behavior portrait corresponding to the sample user.
And the preprocessing module is used for preprocessing the data of the sample long video portrait, the sample short video portrait, the sample user portrait and the sample context behavior portrait to obtain an original characteristic data set.
And the characteristic engineering processing module is used for carrying out characteristic engineering processing on the original characteristic data set to obtain a training data set, wherein the characteristic engineering processing at least comprises characteristic extraction processing and characteristic conversion processing.
In the embodiment of the invention, data preprocessing is carried out on a sample long video image, a sample short video image, a sample user image and a sample context behavior image to obtain an original characteristic data set. And performing characteristic engineering processing such as characteristic extraction processing, characteristic conversion processing and the like on the original characteristic data set to obtain a training data set containing formatting characteristics for training a machine learning model. The machine learning model is trained by the training data set to obtain a recommendation model, and personalized recommendation services for different types of information are provided for the user through the recommendation model, so that the user experience is improved, and the accuracy of information recommendation is improved.
Preferably, in combination with the contents shown in fig. 4, the processing unit 402 for training the machine learning model to obtain the recommendation model includes: the system comprises a dividing module and a training module, wherein the execution principle of each module is as follows.
And the dividing module is used for dividing the training data set into a training set, a verification set and a test set according to a preset dividing proportion.
And the training module is used for training the machine learning model until the machine learning model converges according to the training set, the verification set and the test set and in combination with the click behavior and the exposure behavior of the sample user on the sample long video and the sample short video to obtain a recommendation model.
Preferably, in conjunction with the content shown in fig. 4, the video recommendation system further includes:
and the updating unit is used for updating and optimizing the recommendation model according to the clicking behavior and the exposure behavior of the long video to be recommended and the short video to be recommended of the user.
In summary, embodiments of the present invention provide a video recommendation method and system, where user characteristics and long and short video characteristics obtained by combining long video characteristics and short video characteristics are input into a pre-trained recommendation model, so as to obtain predicted click rates of a long video to be recommended and a short video to be recommended. And sequencing the long video to be recommended and the short video to be recommended according to the predicted click rate, and feeding back a sequencing result to the user, so that the recommendation of different types of information is realized, the user experience is improved, and the accuracy of information recommendation is improved.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, the system or system embodiments are substantially similar to the method embodiments and therefore are described in a relatively simple manner, and reference may be made to some of the descriptions of the method embodiments for related points. The above-described system and system embodiments are only illustrative, wherein the units described as separate parts may or may not be physically separate, and the parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. A method for video recommendation, the method comprising:
acquiring long and short video characteristics and user characteristics of a user, wherein the long and short video characteristics are obtained by combining long video characteristics of a plurality of long videos to be recommended and short video characteristics of a plurality of short videos to be recommended;
inputting the user characteristics and the long and short video characteristics into a preset recommendation model for click rate prediction to obtain the predicted click rate of each long video to be recommended and the predicted click rate of each short video to be recommended, and training a machine learning model according to a training data set obtained in advance by the recommendation model;
and sequencing all the long videos to be recommended and the short videos to be recommended according to the predicted click rate of each long video to be recommended and the predicted click rate of each short video to be recommended, and feeding back sequencing results to the user.
2. The method of claim 1, wherein the obtaining long and short video characteristics and user characteristics of the user comprises:
acquiring historical long and short video characteristics and real-time long and short video characteristics, and fusing the historical long and short video characteristics and the real-time long and short video characteristics to obtain long and short video characteristics;
the method comprises the steps of obtaining historical user characteristics and real-time user characteristics corresponding to a user, and fusing the historical user characteristics and the real-time user characteristics to obtain the user characteristics corresponding to the user.
3. The method of claim 1, wherein obtaining the training data set comprises:
obtaining a sample long video image corresponding to the sample long video and a sample short video portrait corresponding to the sample short video, and obtaining a sample user portrait and a sample context behavior portrait corresponding to the sample user;
performing data preprocessing on the sample long video portrait, the sample short video portrait, the sample user portrait and the sample context behavior portrait to obtain an original characteristic data set;
and carrying out feature engineering processing on the original feature data set to obtain a training data set, wherein the feature engineering processing at least comprises feature extraction processing and feature transformation processing.
4. The method of claim 1, wherein training the machine learning model to obtain the recommendation model according to the pre-acquired training data set comprises:
dividing the training data set into a training set, a verification set and a test set according to a preset division ratio;
and training the machine learning model according to the training set, the verification set and the test set and in combination with the click behavior and the exposure behavior of the sample user on the sample long video and the sample short video until the machine learning model is converged to obtain a recommendation model.
5. The method according to claim 1, wherein all the long videos to be recommended and the short videos to be recommended are sorted, and after the sorting result is fed back to the user, the method further comprises:
and updating and optimizing the recommendation model according to the clicking behavior and the exposure behavior of the user on the long video to be recommended and the short video to be recommended.
6. A video recommendation system, the system comprising:
the device comprises an acquisition unit, a recommendation unit and a recommendation unit, wherein the acquisition unit is used for acquiring long and short video characteristics and user characteristics of a user, and the long and short video characteristics are obtained by combining long video characteristics of a plurality of long videos to be recommended and short video characteristics of a plurality of short videos to be recommended;
the processing unit is used for inputting the user characteristics and the long and short video characteristics into a preset recommendation model for click rate prediction to obtain the predicted click rate of each long video to be recommended and the predicted click rate of each short video to be recommended, and the recommendation model is obtained by training a machine learning model according to a training data set acquired in advance;
and the sequencing unit is used for sequencing all the long videos to be recommended and the short videos to be recommended according to the predicted click rate of each long video to be recommended and the predicted click rate of each short video to be recommended, and feeding back sequencing results to the user.
7. The system of claim 6, wherein the obtaining unit is specifically configured to: the method comprises the steps of obtaining historical long and short video features and real-time long and short video features, fusing the historical long and short video features and the real-time long and short video features to obtain long and short video features, obtaining historical user features and real-time user features corresponding to users, and fusing the historical user features and the real-time user features to obtain user features corresponding to the users.
8. The system of claim 6, wherein the processing unit for obtaining a training data set comprises:
the acquisition module is used for acquiring a sample long video image corresponding to the sample long video and a sample short video portrait corresponding to the sample short video, and acquiring a sample user portrait and a sample context behavior portrait corresponding to a sample user;
the preprocessing module is used for preprocessing the sample long video portrait, the sample short video portrait, the sample user portrait and the sample context behavior portrait to obtain an original characteristic data set;
and the characteristic engineering processing module is used for carrying out characteristic engineering processing on the original characteristic data set to obtain a training data set, wherein the characteristic engineering processing at least comprises characteristic extraction processing and characteristic conversion processing.
9. The system of claim 6, wherein the processing unit for training a machine learning model to obtain a recommendation model comprises:
the dividing module is used for dividing the training data set into a training set, a verification set and a test set according to a preset dividing proportion;
and the training module is used for training the machine learning model according to the training set, the verification set and the test set and combining the click behavior and the exposure behavior of the sample user on the sample long video and the sample short video until the machine learning model is converged to obtain a recommendation model.
10. The system of claim 6, further comprising:
and the updating unit is used for updating and optimizing the recommendation model according to the clicking behavior and the exposure behavior of the user on the long video to be recommended and the short video to be recommended.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010434612.7A CN111339355A (en) | 2020-05-21 | 2020-05-21 | Video recommendation method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010434612.7A CN111339355A (en) | 2020-05-21 | 2020-05-21 | Video recommendation method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111339355A true CN111339355A (en) | 2020-06-26 |
Family
ID=71186514
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010434612.7A Pending CN111339355A (en) | 2020-05-21 | 2020-05-21 | Video recommendation method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111339355A (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111679916A (en) * | 2020-08-11 | 2020-09-18 | 北京搜狐新媒体信息技术有限公司 | Video recommendation method, target service providing terminal, service calling terminal and system |
| CN111898738A (en) * | 2020-07-30 | 2020-11-06 | 北京智能工场科技有限公司 | Mobile terminal user gender prediction method and system based on full-connection neural network |
| CN111914927A (en) * | 2020-07-30 | 2020-11-10 | 北京智能工场科技有限公司 | Mobile app user gender identification method and system for optimizing data imbalance state |
| CN112100444A (en) * | 2020-09-27 | 2020-12-18 | 四川长虹电器股份有限公司 | Search result ordering method and system based on machine learning |
| CN112235603A (en) * | 2020-10-15 | 2021-01-15 | 脸萌有限公司 | Video distribution system, method, computing device and user equipment |
| CN112256918A (en) * | 2020-11-17 | 2021-01-22 | 中国计量大学 | Short video click rate prediction method based on multi-mode dynamic routing |
| CN112395504A (en) * | 2020-12-01 | 2021-02-23 | 中国计量大学 | Short video click rate prediction method based on sequence capsule network |
| CN112669078A (en) * | 2020-12-30 | 2021-04-16 | 上海众源网络有限公司 | Behavior prediction model training method, device, equipment and storage medium |
| CN113038163A (en) * | 2021-03-26 | 2021-06-25 | 百果园技术(新加坡)有限公司 | User experience model training method, short video user experience evaluation method and device |
| CN113204655A (en) * | 2021-07-02 | 2021-08-03 | 北京搜狐新媒体信息技术有限公司 | Multimedia information recommendation method, related device and computer storage medium |
| CN113360770A (en) * | 2021-06-30 | 2021-09-07 | 北京百度网讯科技有限公司 | Content recommendation method, device, equipment and storage medium |
| CN113987260A (en) * | 2021-09-14 | 2022-01-28 | 北京百度网讯科技有限公司 | Video pushing method and device, electronic equipment and storage medium |
| CN114339417A (en) * | 2021-12-30 | 2022-04-12 | 未来电视有限公司 | Video recommendation method, terminal device and readable storage medium |
| CN114579801A (en) * | 2022-04-28 | 2022-06-03 | 深圳市华曦达科技股份有限公司 | Long video recommendation method based on eagle optimization algorithm |
| CN114970955A (en) * | 2022-04-15 | 2022-08-30 | 黑龙江省网络空间研究中心 | Short video heat prediction method and device based on multi-mode pre-training model |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140282709A1 (en) * | 2013-03-14 | 2014-09-18 | Echostar Technologies L.L.C. | Relevant video content pushed to a mobile phone |
| CN106331778A (en) * | 2015-07-06 | 2017-01-11 | 腾讯科技（深圳）有限公司 | Video recommendation method and device |
| CN109753601A (en) * | 2018-11-28 | 2019-05-14 | 北京奇艺世纪科技有限公司 | Recommendation information clicking rate determines method, apparatus and electronic equipment |
| CN110363346A (en) * | 2019-07-12 | 2019-10-22 | 腾讯科技（北京）有限公司 | Clicking rate prediction technique, the training method of prediction model, device and equipment |
-
2020
- 2020-05-21 CN CN202010434612.7A patent/CN111339355A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140282709A1 (en) * | 2013-03-14 | 2014-09-18 | Echostar Technologies L.L.C. | Relevant video content pushed to a mobile phone |
| CN106331778A (en) * | 2015-07-06 | 2017-01-11 | 腾讯科技（深圳）有限公司 | Video recommendation method and device |
| CN109753601A (en) * | 2018-11-28 | 2019-05-14 | 北京奇艺世纪科技有限公司 | Recommendation information clicking rate determines method, apparatus and electronic equipment |
| CN110363346A (en) * | 2019-07-12 | 2019-10-22 | 腾讯科技（北京）有限公司 | Clicking rate prediction technique, the training method of prediction model, device and equipment |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111898738A (en) * | 2020-07-30 | 2020-11-06 | 北京智能工场科技有限公司 | Mobile terminal user gender prediction method and system based on full-connection neural network |
| CN111914927A (en) * | 2020-07-30 | 2020-11-10 | 北京智能工场科技有限公司 | Mobile app user gender identification method and system for optimizing data imbalance state |
| CN111679916A (en) * | 2020-08-11 | 2020-09-18 | 北京搜狐新媒体信息技术有限公司 | Video recommendation method, target service providing terminal, service calling terminal and system |
| CN112100444A (en) * | 2020-09-27 | 2020-12-18 | 四川长虹电器股份有限公司 | Search result ordering method and system based on machine learning |
| CN112235603B (en) * | 2020-10-15 | 2022-04-05 | 脸萌有限公司 | Video distribution system, method, computing device, user equipment and video playing method |
| CN112235603A (en) * | 2020-10-15 | 2021-01-15 | 脸萌有限公司 | Video distribution system, method, computing device and user equipment |
| US11838576B2 (en) | 2020-10-15 | 2023-12-05 | Lemon Inc. | Video distribution system, method, computing device and user equipment |
| CN112256918A (en) * | 2020-11-17 | 2021-01-22 | 中国计量大学 | Short video click rate prediction method based on multi-mode dynamic routing |
| CN112256918B (en) * | 2020-11-17 | 2021-07-06 | 中国计量大学 | Short video click rate prediction method based on multi-mode dynamic routing |
| CN112395504B (en) * | 2020-12-01 | 2021-11-23 | 中国计量大学 | Short video click rate prediction method based on sequence capsule network |
| CN112395504A (en) * | 2020-12-01 | 2021-02-23 | 中国计量大学 | Short video click rate prediction method based on sequence capsule network |
| CN112669078A (en) * | 2020-12-30 | 2021-04-16 | 上海众源网络有限公司 | Behavior prediction model training method, device, equipment and storage medium |
| CN113038163B (en) * | 2021-03-26 | 2023-06-23 | 百果园技术(新加坡)有限公司 | User experience model training method, short video user experience assessment method and device |
| CN113038163A (en) * | 2021-03-26 | 2021-06-25 | 百果园技术(新加坡)有限公司 | User experience model training method, short video user experience evaluation method and device |
| CN113360770A (en) * | 2021-06-30 | 2021-09-07 | 北京百度网讯科技有限公司 | Content recommendation method, device, equipment and storage medium |
| CN113360770B (en) * | 2021-06-30 | 2024-04-12 | 北京百度网讯科技有限公司 | Content recommendation method, device, equipment and storage medium |
| CN113204655A (en) * | 2021-07-02 | 2021-08-03 | 北京搜狐新媒体信息技术有限公司 | Multimedia information recommendation method, related device and computer storage medium |
| CN113204655B (en) * | 2021-07-02 | 2021-11-23 | 北京搜狐新媒体信息技术有限公司 | Multimedia information recommendation method, related device and computer storage medium |
| CN113987260A (en) * | 2021-09-14 | 2022-01-28 | 北京百度网讯科技有限公司 | Video pushing method and device, electronic equipment and storage medium |
| CN114339417A (en) * | 2021-12-30 | 2022-04-12 | 未来电视有限公司 | Video recommendation method, terminal device and readable storage medium |
| CN114970955A (en) * | 2022-04-15 | 2022-08-30 | 黑龙江省网络空间研究中心 | Short video heat prediction method and device based on multi-mode pre-training model |
| CN114970955B (en) * | 2022-04-15 | 2023-12-15 | 黑龙江省网络空间研究中心 | Short video heat prediction method and device based on multi-mode pre-training model |
| CN114579801A (en) * | 2022-04-28 | 2022-06-03 | 深圳市华曦达科技股份有限公司 | Long video recommendation method based on eagle optimization algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111339355A (en) | Video recommendation method and system | |
| CN107491432B (en) | Low-quality article identification method and device based on artificial intelligence, equipment and medium | |
| CN110781321B (en) | Multimedia content recommendation method and device | |
| CN107885796B (en) | Information recommendation method, device and equipment | |
| US8442984B1 (en) | Website quality signal generation | |
| US10692017B2 (en) | Systems and methods for predictive document coding using continuous active machine learning | |
| CN111898032B (en) | Information recommendation method and device based on artificial intelligence, electronic equipment and storage medium | |
| CN111797320B (en) | Data processing method, device, equipment and storage medium | |
| CN112468853B (en) | Television resource recommendation method and device, computer equipment and storage medium | |
| US8024336B2 (en) | System and method for development of search success metrics | |
| CN113626679B (en) | Multimedia resource recommendation method, device and storage medium | |
| CN114610865A (en) | Method, device and equipment for recommending recalled text and storage medium | |
| CN113947131A (en) | Model training method, device, equipment and storage medium | |
| CN113836406A (en) | Information flow recommendation method and device | |
| CN111177564B (en) | Product recommendation method and device | |
| JP2012242844A (en) | Recommendation information generation device and recommendation information generation method | |
| CN112070559A (en) | State acquisition method and device, electronic equipment and storage medium | |
| CN114139046B (en) | Object recommendation method and device, electronic equipment and storage medium | |
| CN112163163B (en) | Multi-algorithm fused information recommendation method, device and equipment | |
| CN113377640A (en) | Method, medium, device and computing equipment for explaining model under business scene | |
| CN113743692A (en) | Business risk assessment method and device, computer equipment and storage medium | |
| CN112231299A (en) | Method and device for dynamically adjusting feature library | |
| CN113157901B (en) | User generated content filtering method and related device | |
| CN114595323B (en) | Portrait construction, recommendation, model training method, apparatus, device and storage medium | |
| CN116823407B (en) | Product information pushing method, device, electronic equipment and computer readable 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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20200806 Address after: 100190, No. 3, No. 2, hospital No. 2, South Road, Beijing Academy of Sciences, 601, 6 Applicant after: BEIJING SOHU NEW POWER INFORMATION TECHNOLOGY Co.,Ltd. Address before: 100190, No. 3, No. 2, hospital No. 2, South Road, Beijing Academy of Sciences, 1201, 12 Applicant before: BEIJING SOHU NEW MEDIA INFORMATION TECHNOLOGY Co.,Ltd. |
|
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200626 |