Video recommendation method and system based on big data

Abstract

The invention provides a video recommendation method and system based on big data, and relates to the technical field of video recommendation, and the method comprises the steps: inputting a user chat record into a video output model, and determining a plurality of discussion videos in the user chat record and a discussion chat record corresponding to each discussion video; based on a plurality of discussion videos in the user chatting records and the discussion chatting records corresponding to each discussion video, using a friend output model to determine a plurality of close discussion friends and closeness degrees of the close discussion friends and the user; constructing a graph structure; processing the graph structure based on a graph neural network model to determine a plurality of recommended videos; according to the method, the video recommendation accuracy can be improved by utilizing the social data of the user, so that the user experience is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of video recommendation, and in particular to a video recommendation method and system based on big data. Background Art

[0002] With the rapid development of Internet technology, watching videos has become an important part of people's daily life. As a bridge connecting users and video content, the performance of the video recommendation system based on big data directly affects the user's viewing experience and the user stickiness of the platform. Traditional video recommendation methods based on big data mainly rely on users' historical behavior data, such as viewing records, search history, etc., and predict videos that users may be interested in through algorithms such as collaborative filtering, content recommendation or hybrid recommendation. However, these methods have some limitations. First, they often ignore the information in the user's social network, such as the interaction and discussion between the user and friends, which contains rich user interests and preferences. Users' video discussions in social networks often involve many aspects of information, such as emotions, opinions and contexts. These complex social signals are difficult to be fully utilized in traditional recommendation systems.

[0003] Therefore, how to use user social data to improve the accuracy of video recommendations and thus improve user experience is a problem that needs to be solved currently. Summary of the invention

[0004] The main technical problem solved by the present invention is how to utilize user social data to improve the accuracy of video recommendations, thereby improving user experience.

[0005] According to a first aspect, the present invention provides a video recommendation method based on big data, comprising: obtaining user chat records; inputting the user chat records into a video output model to determine multiple discussion videos in the user chat records and the discussion chat records corresponding to each discussion video; using a friend output model to determine multiple close discussion friends and the closeness between the multiple close discussion friends and the user based on the multiple discussion videos in the user chat records and the discussion chat records corresponding to each discussion video; obtaining the video browsing records of each close discussion friend and the video browsing records of the user; constructing a graph structure, the graph structure comprising multiple nodes and multiple edges between the multiple nodes, the multiple nodes comprising a user node and multiple close discussion friend nodes, the one user node being a central node, the multiple close discussion friend nodes respectively establishing multiple edges with the one user node, the node feature of the one user node being the video browsing record of the user, the node feature of each close discussion friend node being the video browsing record of each close discussion friend, and the edge feature of the multiple edges being the closeness between the multiple close discussion friends and the user; processing the graph structure based on a graph neural network model to determine multiple recommended videos; and recommending to the user based on the multiple recommended videos.

[0006] Furthermore, the recommendation to users based on the multiple recommended videos includes: performing K-means clustering on the multiple recommended videos to obtain K clusters and cluster centers of each of the K clusters; obtaining the recommended video closest to the cluster center of each cluster to obtain the recommended video corresponding to each cluster; scoring the recommended videos corresponding to each cluster based on a scoring model to obtain a score of the recommended video corresponding to each cluster; arranging the scores of the recommended videos corresponding to each cluster in order from high to low to obtain a display order of the recommended videos corresponding to each cluster; and recommending the recommended videos corresponding to each cluster to users for playback in the display order.

[0007] Furthermore, the recommendation to users based on the multiple recommended videos includes: performing K-means clustering on the multiple recommended videos to obtain K clusters and cluster centers of each of the K clusters; obtaining the recommended video closest to the cluster center of each cluster to obtain the recommended video corresponding to each cluster; determining the closeness of the closely discussed friends corresponding to the recommended videos corresponding to each cluster; arranging the closeness of the closely discussed friends corresponding to the recommended videos corresponding to each cluster in order from high to low to obtain a display order of the recommended videos corresponding to each cluster; and recommending the recommended videos corresponding to each cluster to users for playback in the display order.

[0008] Furthermore, the plurality of recommended videos are selected from the video browsing records of each of the close discussion friends.

[0009] Furthermore, the video output model is a Transformer model, the input of the video output model is the user chat record, and the output of the video output model is multiple discussion videos in the user chat record and the discussion chat record corresponding to each discussion video.

[0010] According to a second aspect, the present invention provides a video recommendation system based on big data, comprising: The first acquisition module is used to acquire user chat records; A video output module, used for inputting the user chat record into a video output model to determine a plurality of discussion videos in the user chat record and a discussion chat record corresponding to each discussion video; A friend output module, configured to determine a plurality of close discussion friends and the closeness between the plurality of close discussion friends and the user using a friend output model based on a plurality of discussion videos in the user chat record and the discussion chat record corresponding to each discussion video; The second acquisition module is used to acquire the video browsing record of each close discussion friend and the video browsing record of the user; A construction module is used to construct a graph structure, wherein the graph structure includes multiple nodes and multiple edges between the multiple nodes, wherein the multiple nodes include a user node and multiple close discussion friend nodes, wherein the one user node is a central node, and the multiple close discussion friend nodes respectively establish multiple edges with the one user node, wherein the node feature of the one user node is the video browsing record of the user, the node feature of each close discussion friend node is the video browsing record of each close discussion friend, and the feature of the edge in the multiple edges is the closeness between the multiple close discussion friends and the user; A recommended video determination module, used to process the graph structure based on a graph neural network model to determine a plurality of recommended videos; The recommendation module is used to make recommendations to users based on the multiple recommended videos.

[0011] Furthermore, the recommendation module is also used to: perform K-means clustering on the multiple recommended videos to obtain K clusters and the cluster center of each cluster in the K clusters; obtain the recommended video closest to the cluster center of each cluster to obtain the recommended video corresponding to each cluster; score the recommended videos corresponding to each cluster based on the scoring model to obtain the score of the recommended video corresponding to each cluster; arrange the scores of the recommended videos corresponding to each cluster in descending order to obtain the display order of the recommended videos corresponding to each cluster; and recommend the recommended videos corresponding to each cluster to users for playback in the display order.

[0012] Furthermore, the recommendation module is also used to: perform K-means clustering on the multiple recommended videos to obtain K clusters and the cluster center of each cluster in the K clusters; obtain the recommended video closest to the cluster center of each cluster to obtain the recommended video corresponding to each cluster; determine the closeness of the closely discussed friends corresponding to the recommended videos corresponding to each cluster; arrange the closeness of the closely discussed friends corresponding to the recommended videos corresponding to each cluster in order from high to low to obtain the display order of the recommended videos corresponding to each cluster; and recommend the recommended videos corresponding to each cluster to users for playback in the display order.

[0013] Furthermore, the plurality of recommended videos are selected from the video browsing records of each of the close discussion friends.

[0014] Furthermore, the video output model is a Transformer model, the input of the video output model is the user chat record, and the output of the video output model is multiple discussion videos in the user chat record and the discussion chat record corresponding to each discussion video.

[0015] The present invention provides a video recommendation method and system based on big data, which includes obtaining user chat records; inputting the user chat records into a video output model to determine multiple discussion videos in the user chat records and the discussion chat records corresponding to each discussion video; using a friend output model based on the multiple discussion videos in the user chat records and the discussion chat records corresponding to each discussion video to determine multiple close discussion friends and the closeness between the multiple close discussion friends and the user; obtaining the video browsing records of each close discussion friend and the video browsing records of the user; constructing a graph structure, wherein the graph structure includes multiple nodes and multiple edges between the multiple nodes, and the multiple nodes include A user node and multiple close discussion friend nodes, the one user node is a central node, the multiple close discussion friend nodes respectively establish multiple edges with the one user node, the node feature of the one user node is the video browsing record of the user, the node feature of each close discussion friend node is the video browsing record of each close discussion friend, and the edge feature of the multiple edges is the closeness between the multiple close discussion friends and the user; the graph structure is processed based on a graph neural network model to determine multiple recommended videos; recommendations are made to users based on the multiple recommended videos. The method can utilize user social data to improve the accuracy of video recommendations, thereby improving user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic diagram of a process of a video recommendation method based on big data provided by an embodiment of the present invention; Figure 2A schematic diagram of a process of recommending a user based on multiple recommended videos provided by an embodiment of the present invention; Figure 3 A schematic diagram of another process of recommending a user based on multiple recommended videos provided by an embodiment of the present invention; Figure 4 A schematic diagram of a video recommendation system based on big data provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0017] In an embodiment of the present invention, there is provided Figure 1 A video recommendation method based on big data is shown, and the video recommendation method based on big data includes steps S1 to S7: Step S1, obtaining user chat records.

[0018] User chat records refer to the conversation content generated when users communicate with others in chat applications. User chat records include text, video, images, etc.

[0019] Step S2: input the user chat record into a video output model to determine multiple discussion videos in the user chat record and the discussion chat record corresponding to each discussion video.

[0020] The video output model is a Transformer model, the input of the video output model is the user chat record, and the output of the video output model is multiple discussion videos in the user chat record and the discussion chat record corresponding to each discussion video.

[0021] The Transformer model consists of an encoder and a decoder. The encoder is used to learn the representation of the input sequence, which includes a self-attention mechanism and a feed-forward neural network. Based on the encoder, the decoder introduces an additional multi-head attention mechanism to decode the encoder output and generate the target sequence. The Transformer model can be used to process videos in user chat records and better capture the relationship in the time series of the video.

[0022] The multiple discussion videos in the user chat record are video contents discussed with friends in the user chat record. For example, if a user sends a landscape video to a friend and has a text discussion with the friend, the landscape video is a discussion video.

[0023] The discussion chat record corresponding to each discussion video refers to the text content in the user chat record associated with the specific discussion video. For example, for a discussion video, its corresponding discussion chat record includes the relevant conversation content related to the video topic in the chat record. For example, a user sends a landscape video to a friend and has a corresponding text discussion with the friend. As an example, the discussion chat record corresponding to the landscape video is "This lake is so beautiful."

[0024] The Transformer model can simultaneously focus on information at different positions in the input sequence through the self-attention mechanism, and performs well in handling long-distance dependencies. This enables the model to identify video-related keywords or features in the user chat history, and thus determine the multiple discussion videos involved in the user chat history. The Transformer model combines visual and language information and can perform visual-language interaction to identify video content or related descriptions in the user chat history, and then determine the multiple discussion videos in the user chat history. The Transformer model can model and understand the context in the user chat history to better grasp the user's discussion content. By capturing the association between each word and its surrounding words, the model can more accurately determine the context of the discussion video, and then determine the multiple discussion videos in the user chat history. The Transformer model can simultaneously focus on information at different positions and contents through the multi-head attention mechanism. This enables the model to accurately determine the discussion chat record corresponding to each discussion video in the user chat history, thereby distinguishing different video discussion contents.

[0025] In some embodiments, the video output model includes a chat record screening layer, a discussion video output layer, and a chat record determination layer. The chat record screening layer, the discussion video output layer, and the chat record determination layer all include a Transformer structure. The input of the chat record screening layer is the user chat record, the output of the chat record screening layer is the preliminarily screened discussion chat record related to the video, the input of the discussion video output layer is the preliminarily screened discussion chat record related to the video, the output of the discussion video output layer is multiple discussion videos, the input of the chat record determination layer is multiple discussion videos and preliminarily screened discussion chat records related to the video, and the output of the chat record determination layer is the discussion chat record corresponding to each discussion video. Through the chat record screening layer, the discussion chat record related to the video can be preliminarily screened out from the user's chat record, and the discussion video output layer can process the preliminarily screened discussion chat record related to the video and output multiple discussion videos in the user's chat record. The chat record determination layer can clearly determine the discussion chat record corresponding to each discussion video and establish a corresponding relationship between the video and the chat record. By dividing the video output model into a chat record initial screening layer, a discussion video output layer and a chat record confirmation layer, it is possible to quickly screen relevant chat records, extract users' existing video resources, and establish a correspondence between videos and chat records, thereby improving processing efficiency and accuracy.

[0026] Step S3, based on the multiple discussion videos in the user chat record and the discussion chat record corresponding to each discussion video, use the friend output model to determine multiple close discussion friends and the closeness between the multiple close discussion friends and the user.

[0027] The friend output model is a Transformer model, the input of the friend output model is multiple discussion videos in the user chat record and the discussion chat record corresponding to each discussion video, and the output of the friend output model is multiple close discussion friends and the closeness between the multiple close discussion friends and the user.

[0028] Multiple close discussion friends refer to the group of friends with whom the user interacts with video content most frequently and discusses the most, determined through Transformer model analysis and inference based on the user's chat records and discussion video content.

[0029] The closeness between a close discussion friend and a user refers to the intimacy of the close discussion friend's interaction with the user in video content. For example, if a user often discusses video content with certain friends in chat records, and these discussion videos and chat records are input into the Transformer model, the model can analyze and determine the group of friends who interact with the user's video discussions most frequently. These friends can be identified as close discussion friends. At the same time, the Transformer model will also determine the closeness between multiple close discussion friends and the user based on the user's chat records and discussion video content. The greater the closeness, the more frequent the discussion between the close discussion friend and the user on video content.

[0030] By analyzing the discussion videos and corresponding discussion chat records in the user's chat history, the Transformer model can capture rich contextual information, including conversation content, sentiment tendencies, common topics, etc., thereby inferring the friends who have close discussions with the user and outputting the degree of closeness between the close discussion friends and the user.

[0031] Step S4, obtaining the video browsing record of each close discussion friend and the video browsing record of the user.

[0032] The video browsing history of close discussion friends is the record generated when close discussion friends watch videos on the platform. The video browsing history of close discussion friends includes information such as the video content watched, viewing duration, and viewing time.

[0033] A user's video browsing history is a record generated when a user watches videos on the platform. The user's video browsing history includes information such as the video content watched, viewing duration, and viewing time.

[0034] Step S5, constructing a graph structure, wherein the graph structure includes multiple nodes and multiple edges between the multiple nodes, the multiple nodes include a user node and multiple close discussion friend nodes, the one user node is a central node, the multiple close discussion friend nodes respectively establish multiple edges with the one user node, the node feature of the one user node is the video browsing record of the user, the node feature of each close discussion friend node is the video browsing record of each close discussion friend, and the feature of the edge in the multiple edges is the closeness between the multiple close discussion friends and the user.

[0035] The graph structure is a data structure consisting of nodes and edges. A node includes node features, which are attributes or features of a node and can be used to describe the nature or information of a node. An edge is used to describe the relationship between two nodes. The feature of an edge is the degree of closeness between the multiple close discussion friends and the user.

[0036] Step S6: Process the graph structure based on the graph neural network model to determine multiple recommended videos.

[0037] The graph neural network model includes a graph neural network (GNN) and a fully connected layer. The graph neural network is a neural network that directly acts on graph structure data. The input of the graph neural network model is the graph structure, and the output of the graph neural network model is multiple recommended videos.

[0038] Recommended videos are videos output by the graph neural network model and recommended to users.

[0039] By constructing a graph structure, the relationship network between users and their close friends can be clearly reflected. This relationship information is very important for personalized recommendations because users' interests and preferences are often influenced by their close friends. Taking the video browsing records of user nodes and close friends' nodes as node features and the closeness as edge features can make more full use of data information. This helps the model better understand the interaction and influence between users and friends and improve the accuracy of recommendations. Processing graph structured data based on the graph neural network model can effectively learn the complex relationships and information transmission between nodes, thereby more accurately mining users' interests and social relationships. Compared with traditional recommendation algorithms, graph neural networks have better representation and learning capabilities when processing graph data.

[0040] Step S7: recommending videos to users based on the multiple recommended videos.

[0041] In some embodiments, the plurality of recommended videos may be screened by a scoring model before being recommended to the user. Figure 2 A schematic diagram of a process of recommending to a user based on multiple recommended videos provided by an embodiment of the present invention, wherein the process of recommending to a user based on multiple recommended videos includes steps S21 to S25: Step S21 , performing K-means clustering on the plurality of recommended videos to obtain K clusters and a cluster center of each of the K clusters.

[0042] In some embodiments, the features of the multiple recommended videos can be extracted through a feature extraction model to obtain color histograms of the multiple recommended videos, optical flows of the multiple recommended videos, and Mel-frequency cepstral coefficients of the multiple recommended videos. Based on the color histograms of the multiple recommended videos, the optical flows of the multiple recommended videos, the Mel-frequency cepstral coefficients of the multiple recommended videos, and the pre-set K value, a K-means clustering algorithm is used to obtain K clusters and cluster centers of each cluster in the K clusters, and each cluster represents a class of similar recommended videos. As an example, specifically: K recommended videos are randomly selected as initial cluster centers. For each recommended video, the distance between each recommended video and all initial cluster centers is calculated, and each recommended video is assigned to the nearest cluster center to form K clusters. For each cluster, the average value of the feature vectors of all videos in the cluster is calculated, and this average value is used as the new cluster center. The above steps are repeated until the cluster center reaches a predetermined number of iterations, and the clustering ends.

[0043] The feature extraction model is a Transformer model, the input of the feature extraction model is multiple recommended videos, and the output of the feature extraction model is color histograms of the multiple recommended videos, optical flows of the multiple recommended videos, and Mel-frequency cepstral coefficients of the multiple recommended videos.

[0044] The color histogram of a video is used to describe the statistical characteristics of color distribution in a video. The color histogram divides the colors in the video into several intervals according to certain rules, and then counts the number of pixels in each color interval to finally form a histogram.

[0045] The optical flow of a video is used to describe the movement of pixels in an image over time. By calculating the optical flow, we can obtain the movement trajectory and speed information of objects in the video.

[0046] The Mel-frequency cepstral coefficients of the video are used to represent the coefficients of the spectral characteristics of the audio signal.

[0047] In some embodiments, the value of K can be determined by a preset relationship table between the value of K and the average video browsing time of the user per day. The longer the average video browsing time of the user per day, the larger the value of K. The preset relationship table between the value of K and the average video browsing time of the user per day is artificially constructed in advance.

[0048] Step S22, obtaining the recommended video closest to the cluster center of each cluster to obtain the recommended video corresponding to each cluster.

[0049] For each cluster, calculate the distance between each recommended video feature vector and the cluster center of the cluster. For each cluster, find the recommended video closest to the cluster center, and this video will be the recommended video corresponding to each cluster.

[0050] Step S23, scoring the recommended videos corresponding to each cluster based on the scoring model to obtain the scores of the recommended videos corresponding to each cluster.

[0051] The scoring model is a Transformer model, the input of which is a plurality of recommended videos, and the output of which is the color histograms of the plurality of recommended videos, the optical flows of the plurality of recommended videos, and the Mel-frequency cepstral coefficients of the plurality of recommended videos. The scoring model is a model for evaluating the quality of the recommended videos corresponding to each cluster. The scoring model can process the recommended videos corresponding to each cluster and give higher scores to videos with more refined video images, higher definition, and more attractive video content. The higher the score of the recommended videos corresponding to each cluster, the better the quality of the video.

[0052] Step S24, arranging the scores of the recommended videos corresponding to each cluster in descending order to obtain a display order of the recommended videos corresponding to each cluster.

[0053] The recommended videos corresponding to each cluster are sorted from high to low according to the scores to obtain the display order of the recommended videos corresponding to each cluster.

[0054] Step S25: recommending the recommended videos corresponding to each cluster to the user for playback according to the display order of the recommended videos.

[0055] After the display order of the recommended videos corresponding to each cluster is obtained, they are recommended to the user for playback.

[0056] In some embodiments, the plurality of recommended videos may be screened according to the closeness of the close discussion friends corresponding to the recommended videos before being recommended to the user. Figure 3 Another flowchart of recommending a user based on multiple recommended videos provided by an embodiment of the present invention includes steps S31 to S35: Step S31, performing K-means clustering on the plurality of recommended videos to obtain K clusters and a cluster center of each of the K clusters.

[0057] For step S31, please refer to step S21, which will not be repeated here.

[0058] Step S32, obtaining the recommended video closest to the cluster center of each cluster to obtain the recommended video corresponding to each cluster.

[0059] For step S32, please refer to step S22, which will not be repeated here.

[0060] Step S33, determining the closeness of the closely discussed friends corresponding to the recommended videos corresponding to each cluster.

[0061] After the recommended videos corresponding to each cluster are determined, the closeness of the close discussion friends of the recommended videos corresponding to each cluster is obtained based on the multiple close discussion friends determined in step S3 and the closeness between the multiple close discussion friends and the user. The greater the closeness of the close discussion friends of the recommended videos corresponding to each cluster, the closer the relationship between the user and the close discussion friends in discussing the videos is, and the video can be recommended first.

[0062] Step S34, arranging the closeness of the close discussion friends corresponding to the recommended videos corresponding to each cluster in descending order to obtain a display order of the recommended videos corresponding to each cluster.

[0063] The recommended videos corresponding to each cluster are sorted from high to low according to the closeness of the close discussion friends corresponding to the recommended videos corresponding to each cluster to obtain a display order of the recommended videos corresponding to each cluster.

[0064] Step S35: recommending the recommended videos corresponding to each cluster to the user for playback according to the display order of the recommended videos.

[0065] After the display order of the recommended videos corresponding to each cluster is obtained, they are recommended to the user for playback.

[0066] Based on the same inventive concept, Figure 4 A schematic diagram of a video recommendation system based on big data provided by an embodiment of the present invention, wherein the video recommendation system based on big data includes: A first acquisition module 41 is used to acquire user chat records; The video output module 42 is used to input the user chat record into the video output model to determine multiple discussion videos in the user chat record and the discussion chat record corresponding to each discussion video; A friend output module 43, configured to determine a plurality of close discussion friends and the closeness between the plurality of close discussion friends and the user using a friend output model based on a plurality of discussion videos in the user chat record and the discussion chat record corresponding to each discussion video; The second acquisition module 44 is used to acquire the video browsing record of each close discussion friend and the video browsing record of the user; A construction module 45 is used to construct a graph structure, wherein the graph structure includes multiple nodes and multiple edges between the multiple nodes, wherein the multiple nodes include a user node and multiple close discussion friend nodes, wherein the one user node is a central node, and the multiple close discussion friend nodes respectively establish multiple edges with the one user node, wherein the node feature of the one user node is the video browsing record of the user, the node feature of each close discussion friend node is the video browsing record of each close discussion friend, and the feature of the edge in the multiple edges is the closeness between the multiple close discussion friends and the user; A recommended video determination module 46, configured to process the graph structure based on a graph neural network model to determine a plurality of recommended videos; The recommendation module 47 is used to make recommendations to users based on the multiple recommended videos.

Claims

1. A video recommendation method based on big data, characterized in that: include: Get user chat history; Inputting the user chat record into a video output model to determine a plurality of discussion videos in the user chat record and a discussion chat record corresponding to each discussion video; Determine multiple close discussion friends and the closeness between the multiple close discussion friends and the user using a friend output model based on multiple discussion videos in the user chat record and the discussion chat record corresponding to each discussion video; Get the video browsing history of each close discussion friend and the user's video browsing history; Constructing a graph structure, wherein the graph structure includes a plurality of nodes and a plurality of edges between the plurality of nodes, wherein the plurality of nodes include a user node and a plurality of close discussion friend nodes, wherein the user node is a central node, and the plurality of close discussion friend nodes respectively establish a plurality of edges with the user node, wherein a node feature of the user node is a video browsing record of the user, a node feature of each close discussion friend node is a video browsing record of each close discussion friend, and a feature of an edge in the plurality of edges is a degree of closeness between the plurality of close discussion friends and the user; Processing the graph structure based on a graph neural network model to determine a plurality of recommended videos; Recommendations are made to users based on the multiple recommended videos.

2. The video recommendation method based on big data as claimed in claim 1, characterized in that: The recommending to the user based on the multiple recommended videos comprises: Performing K-means clustering on the multiple recommended videos to obtain K clusters and a cluster center of each of the K clusters; Get the recommended video closest to the cluster center of each cluster to get the recommended video corresponding to each cluster; The recommended videos corresponding to each cluster are scored based on the scoring model to obtain the scores of the recommended videos corresponding to each cluster; Arranging the scores of the recommended videos corresponding to each cluster in descending order to obtain a display order of the recommended videos corresponding to each cluster; The recommended videos corresponding to each cluster are recommended to the user for playback according to the display order of the recommended videos.

3. The video recommendation method based on big data as claimed in claim 1, characterized in that: The recommending to the user based on the multiple recommended videos comprises: Performing K-means clustering on the multiple recommended videos to obtain K clusters and a cluster center of each of the K clusters; Get the recommended video closest to the cluster center of each cluster to get the recommended video corresponding to each cluster; Determine the closeness of the closely discussed friends corresponding to the recommended video corresponding to each cluster; Arrange the closeness of the closely discussed friends corresponding to the recommended videos corresponding to each cluster in descending order to obtain a display order of the recommended videos corresponding to each cluster; The recommended videos corresponding to each cluster are recommended to the user for playback according to the display order of the recommended videos.

4. The video recommendation method based on big data as claimed in claim 1, characterized in that: The multiple recommended videos are selected from the video browsing records of each of the close discussion friends.

5. The video recommendation method based on big data as claimed in claim 1, characterized in that: The video output model is a Transformer model, the input of the video output model is the user chat record, and the output of the video output model is multiple discussion videos in the user chat record and the discussion chat record corresponding to each discussion video.

6. A video recommendation system based on big data, characterized in that: include: The first acquisition module is used to acquire user chat records; A video output module, used for inputting the user chat record into a video output model to determine a plurality of discussion videos in the user chat record and a discussion chat record corresponding to each discussion video; A friend output module, configured to determine a plurality of close discussion friends and the closeness between the plurality of close discussion friends and the user using a friend output model based on a plurality of discussion videos in the user chat record and the discussion chat record corresponding to each discussion video; The second acquisition module is used to acquire the video browsing record of each close discussion friend and the video browsing record of the user; A construction module is used to construct a graph structure, wherein the graph structure includes multiple nodes and multiple edges between the multiple nodes, wherein the multiple nodes include a user node and multiple close discussion friend nodes, wherein the one user node is a central node, and the multiple close discussion friend nodes respectively establish multiple edges with the one user node, wherein the node feature of the one user node is the video browsing record of the user, the node feature of each close discussion friend node is the video browsing record of each close discussion friend, and the feature of the edge in the multiple edges is the closeness between the multiple close discussion friends and the user; A recommended video determination module, used to process the graph structure based on a graph neural network model to determine a plurality of recommended videos; The recommendation module is used to make recommendations to users based on the multiple recommended videos.

7. The video recommendation system based on big data as claimed in claim 6, characterized in that: The recommendation module is also used to: Performing K-means clustering on the multiple recommended videos to obtain K clusters and a cluster center of each of the K clusters; Get the recommended video closest to the cluster center of each cluster to get the recommended video corresponding to each cluster; The recommended videos corresponding to each cluster are scored based on the scoring model to obtain the scores of the recommended videos corresponding to each cluster; Arranging the scores of the recommended videos corresponding to each cluster in descending order to obtain a display order of the recommended videos corresponding to each cluster; The recommended videos corresponding to each cluster are recommended to the user for playback according to the display order of the recommended videos.

8. The video recommendation system based on big data as claimed in claim 6, characterized in that: The recommendation module is further used for: performing K-means clustering on the plurality of recommended videos to obtain K clusters and a cluster center of each of the K clusters; Get the recommended video closest to the cluster center of each cluster to get the recommended video corresponding to each cluster; Determine the closeness of the closely discussed friends corresponding to the recommended video corresponding to each cluster; Arrange the closeness of the closely discussed friends corresponding to the recommended videos corresponding to each cluster in descending order to obtain a display order of the recommended videos corresponding to each cluster; The recommended videos corresponding to each cluster are recommended to the user for playback according to the display order of the recommended videos.

9. The video recommendation system based on big data as claimed in claim 6, characterized in that: The multiple recommended videos are selected from the video browsing records of each of the close discussion friends.

10. The video recommendation system based on big data as claimed in claim 6, characterized in that: The video output model is a Transformer model, the input of the video output model is the user chat record, and the output of the video output model is multiple discussion videos in the user chat record and the discussion chat record corresponding to each discussion video.

Images