Media data analysis method and system based on artificial intelligence

Abstract

The invention belongs to the technical field of data analysis, and discloses a media data analysis method and system based on artificial intelligence. The method comprises the following steps: establishing a media data crawling engine, a media data processing engine and a media data analysis engine in a media data analysis platform; using a media data crawling engine to generate real-time data crawling strategies of different media platforms, and crawling a plurality of pieces of real-time multi-mode media data; using a media data processing engine to perform data processing on the plurality of pieces of real-time multi-mode media data to obtain a plurality of real-time media data clusters and real-time cluster portraits thereof; and performing multi-dimensional analysis by using a media data analysis engine according to the plurality of real-time media data clusters and the real-time cluster portraits thereof to obtain a real-time media data analysis report. According to the method, the problems of hysteresis and one-sidedness of data acquisition, insufficient multi-modal data processing capability and single analysis dimension in the prior art are solved.

Description

Technical Field

[0001] The present invention belongs to the field of data analysis technology, and specifically relates to a media data analysis method and system based on artificial intelligence. Background Art

[0002] With the rapid development of the internet and social media, the speed and breadth of information dissemination are increasing exponentially. Media platforms (such as news websites, social media, and video platforms) generate massive amounts of multimodal data daily, including text, images, videos, user information, and interaction data. This data contains a wealth of information on social hot spots, public sentiment, and communication patterns, making it crucial for media organizations, governments, and businesses to understand public opinion, formulate strategies, and optimize communication.

[0003] However, traditional media data analysis methods often face the following challenges: 1) Data collection is lagging and incomplete: Most existing methods rely on scheduled crawling or offline batch processing, which cannot capture the real-time dynamics of information dissemination. Data crawling is often limited to pre-built databases on the internet, ignoring many emerging, vertical or regional social media, forums, short video media platforms, etc., resulting in a lack of comprehensiveness in analysis results and the potential omission of important information sources or user groups. In addition, media data crawling strategies are mostly fixed rules or simple heuristics, which are difficult to adapt to the complex and changing page structures, anti-crawl mechanisms, and real-time changes in data update priorities on different platforms. 2) Insufficient multimodal data processing capabilities: Many existing technologies can only process a single modality (e.g., plain text) or only superficially process non-textual modalities (e.g., images, videos) (e.g., extracting only titles or simple tags). Text, images, videos, user information, and interactive data are often processed separately, making cross-modal correlation analysis and information fusion impossible. 3) Single analysis dimension: Most existing technologies analyze media data based on a single dimension, such as data volume, theme, or user group. This results in low value of media data analysis results and a lack of systematic integrated analysis results from multiple dimensions. Summary of the Invention

[0004] In order to solve the problems of data collection lag and one-sidedness, insufficient multimodal data processing capabilities and single analysis dimension in the existing technology, the purpose of the present invention is to provide a media data analysis method and system based on artificial intelligence.

[0005] The technical solution adopted in the present invention is: A media data analysis method based on artificial intelligence comprises the following steps: Based on artificial intelligence algorithms, build media data crawling engines, media data processing engines, and media data analysis engines in the media data analysis platform; Use the media data crawling engine to generate real-time data crawling strategies for different media platforms, and crawl a number of real-time multimodal media data based on the real-time data crawling strategies; Using a media data processing engine, processing a number of real-time multimodal media data to obtain a number of real-time media data clusters and real-time cluster portraits; Based on a number of real-time media data clusters and their real-time cluster portraits, a media data analysis engine is used to perform multi-dimensional analysis to obtain a real-time media data analysis report.

[0006] Furthermore, the real-time multimodal media data includes real-time text data, real-time image data, real-time video data, real-time user information data and real-time interactive data of the media platform.

[0007] Furthermore, the media data crawling engine is provided with a media data crawling interface, a media data crawling executor, and a data crawling strategy generator connected in sequence. The media data crawling interface is respectively connected to a plurality of external media platforms. The data crawling strategy generator is provided with a data crawling strategy generation model. The data crawling strategy generation model is constructed based on the MAPGRPO algorithm, and the data crawling strategy generation model includes a platform-level data crawling strategy generation layer, a coordination layer, and a server-level data crawling strategy generation layer connected in sequence. The platform-level data crawling strategy generation layer is provided with a first multi-optimization target set and a platform-level intelligent agent, the coordination layer is provided with a coordination intelligent agent and a multi-target conflict resolution mechanism, and the server-level data crawling strategy generation layer is provided with a second multi-optimization target set and several parallel server-level intelligent agents. Furthermore, the media data processing engine is provided with a media data preprocessor, a multimodal feature extractor, a multimodal feature spatiotemporal aligner, a spatiotemporal-aware incremental clusterer, and a clustering cluster portrait generator, which are connected in sequence; The multimodal feature extractor is provided with a multimodal feature extraction model, which is constructed based on the BERT-ResNet-CNN-RNN-Wav2Vec2.0-AE algorithm, and the multimodal feature extraction model includes a text feature extraction module constructed based on the BERT algorithm, an image feature extraction module constructed based on the ResNet algorithm, a video feature extraction module constructed based on the CNN-RNN-Wav2Vec2.0 algorithm, a user feature extraction module constructed based on the AE algorithm, and a propagation feature extraction module constructed based on the AE algorithm; The spatiotemporal-aware incremental clusterer is provided with a spatiotemporal-aware incremental clustering model, which is constructed based on the STICM algorithm. The spatiotemporal-aware incremental clustering model is provided with a feature clustering module constructed based on the ST-DBSCAN algorithm, a new data processing module constructed based on the IP algorithm, and a cluster management module constructed based on the DCM algorithm. The clustering portrait generator is provided with a clustering portrait generation model, which is constructed based on the GCN-DualGAN algorithm, and the clustering portrait generation model includes a clustering relationship construction module constructed based on the GCN algorithm and a clustering portrait generation module constructed based on the DualGAN algorithm, which are connected in sequence.

[0008] Furthermore, the media data analysis engine is provided with a media data analysis model, which includes a feature selection module based on the CA algorithm, a multi-dimensional analysis module based on the MDA algorithm, and a report generation module based on the cGAN algorithm, which are connected in sequence. The multi-dimensional analysis module is provided with several parallel dimensional analysis channels. The dimensional analysis channels include the content topic evolution analysis channel built based on the LDA-DTM algorithm, the sentiment distribution analysis channel built based on the VADER-LSTM algorithm, the key communication analysis channel built based on the SNA algorithm, the user portrait analysis channel built based on the GMM-SVM algorithm, the spatiotemporal distribution analysis channel built based on the STGNN algorithm, and the media content analysis channel built based on the VGG algorithm.

[0009] Furthermore, a media data crawling engine is used to generate real-time data crawling strategies for different media platforms, and a plurality of real-time multimodal media data are crawled according to the real-time data crawling strategies, including the following steps: Use the media data crawling interface of the media data crawling engine to collect basic information from different media platforms and input the basic information into the data crawling strategy generator; Based on the basic information, use the data crawling strategy generator of the media data crawling engine to generate strategies and obtain real-time data crawling strategies; According to the real-time data crawling strategy, the media data crawling executor of the media data crawling engine is used to crawl a number of real-time multimodal media data of the corresponding media platform.

[0010] Furthermore, a media data processing engine is used to process a plurality of real-time multimodal media data to obtain a plurality of real-time media data clusters and real-time cluster portraits, including the following steps: Using a media data preprocessor of a media data processing engine, preprocessing a plurality of real-time multimodal media data from different media platforms to obtain a plurality of preprocessed real-time multimodal media data; Using a multimodal feature extractor of a media data processing engine, extracting real-time multimodal feature vectors of a plurality of pre-processed real-time multimodal media data; Using a multimodal feature spatiotemporal aligner of a media data processing engine, performing spatiotemporal alignment on a plurality of real-time multimodal feature vectors to obtain a real-time spatiotemporal alignment matrix including the plurality of real-time multimodal feature vectors; Using the time-space-aware incremental clusterer of the media data processing engine, clustering a number of real-time multimodal feature vectors in the real-time time-space alignment matrix to obtain a number of real-time media data clusters; The clustering portrait generator of the media data processing engine is used to generate a real-time cluster portrait corresponding to each real-time media data cluster.

[0011] Furthermore, multi-dimensional analysis includes content theme evolution analysis, sentiment distribution analysis, key communication analysis, user portrait analysis, spatiotemporal distribution analysis, and media content analysis.

[0012] Furthermore, based on the real-time media data clusters and their real-time cluster portraits, a media data analysis engine is used to perform multi-dimensional analysis to obtain a real-time media data analysis report, including the following steps: Input the real-time media data clusters and their real-time cluster portraits into the media data analysis engine, and use the feature selection module of the media data analysis engine to select real-time key features of different dimensions; Input the real-time key features of different dimensions into the corresponding dimension analysis channel of the multi-dimensional analysis module of the media data analysis engine for analysis to obtain the corresponding real-time dimension analysis results; Integrate the real-time dimensional analysis results input by all dimensional analysis channels to obtain the real-time media data clustering clusters and their real-time cluster portraits corresponding to the real-time media data analysis results; A real-time media data analysis report is generated using a report generation module of a media data analysis engine according to the real-time media data analysis results of all real-time media data clusters.

[0013] A media data analysis system based on artificial intelligence is used to implement a media data analysis method. The system is set on a media data analysis platform, and the system includes a media data crawling engine, a media data processing engine and a media data analysis engine connected in sequence. The media data crawling engine is respectively connected to several external media platforms.

[0014] The beneficial effects of the present invention are: The present invention provides an artificial intelligence-based media data analysis method and system, which dynamically generates real-time strategies through an artificial intelligence-driven media data crawling engine and efficiently captures data. Combined with the rapid response of the processing and analysis engine, the entire analysis process can keep up with the real-time dynamics of information dissemination, greatly shortening the time from data generation to analysis result output, solving the core problem of lagging analysis in existing technologies, and being able to instantly capture the rise and evolution of emergencies and hot topics, providing valuable first-time insights for public opinion monitoring, crisis warning, market response, etc. The media data crawling engine can generate strategies for different media platforms, effectively covering mainstream and emerging platforms, avoiding information blind spots, making the analysis results more representative, and can dynamically adjust the crawling strategy according to data feedback and goals, automatically adapting to platform changes and hot spot shifts, improving the efficiency and pertinence of data acquisition, and enhancing data coverage. comprehensiveness and intelligence; the media data processing engine can uniformly process multiple modal data such as text, images, and videos, and identify the relationships between them, providing a more three-dimensional and complete information view, and can effectively extract and analyze the spatiotemporal attributes of data, revealing the geographical distribution and temporal patterns of information dissemination, and providing a new dimension for understanding the dissemination mechanism. The feature selection, multi-dimensional analysis and other modules of the media data processing engine are all driven by artificial intelligence, which can automatically identify key information, conduct more in-depth and objective analysis, and reduce human intervention and subjective bias; the media data analysis engine conducts media data analysis from multiple dimensions such as content theme evolution analysis, sentiment distribution analysis, key communication analysis, user portrait analysis, spatiotemporal distribution analysis, and media content analysis, which improves the value of media data analysis results and obtains a systematic integrated analysis report, improving the convenience and efficiency of media data analysis.

[0015] Other beneficial effects of the present invention will be further described in the specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a flowchart of the media data analysis method based on artificial intelligence in the present invention.

[0017] Figure 2 It is a structural block diagram of the media data analysis system based on artificial intelligence in the present invention. DETAILED DESCRIPTION

[0018] The present invention will be further explained below with reference to the accompanying drawings and specific embodiments.

[0019] Example 1: like Figure 1 As shown, this embodiment provides a media data analysis method based on artificial intelligence, including the following steps: S1: Based on artificial intelligence algorithms, build media data crawling engine, media data processing engine and media data analysis engine in the media data analysis platform; The media data crawling engine is provided with a media data crawling interface, a media data crawling executor, and a data crawling strategy generator, which are connected in sequence. The media data crawling interface is connected to several external media platforms (such as Weibo, WeChat official accounts, Douyin, Bilibili, Zhihu, news websites, etc.). The data crawling strategy generator is provided with a data crawling strategy generation model. The data crawling strategy generation model is constructed based on the Multi-Agent Parallel Group Relative Policy Optimization (MAPGRPO) algorithm, and the data crawling strategy generation model includes a platform-level data crawling strategy generation layer, a coordination layer, and a server-level data crawling strategy generation layer connected in sequence. The platform-level data crawling strategy generation layer is provided with a first multi-optimization target set (for example: total data volume, data quality, unit time cost, compliance risk index) and a platform-level agent, which is responsible for formulating global collection priorities, resource allocation strategies, bypassing anti-crawler mechanisms, etc. The coordination layer is provided with a coordination agent and a multi-target conflict resolution mechanism to deal with conflicts between platform-level and platform-level targets (for example, the system wants to be fast, but the platform The server-level data crawling strategy generation layer is equipped with a second set of optimization objectives (for example, the success rate of data acquisition on a specific platform, request latency, and detection probability) and several parallel target platform-level agents. Each agent corresponds to one or a category of target media platforms and is responsible for generating specific request frequencies, User-Agent rotation, proxy IP usage, Cookies management, session persistence, and bypass sub-strategies for the platform's specific anti-crawling mechanism (such as processing after verification code recognition is triggered, simulation of sliding verification, response to human-machine verification, etc.); The media data processing engine is provided with a media data preprocessor, a multimodal feature extractor, a multimodal feature spatiotemporal aligner, a spatiotemporal perception incremental clusterer, and a clustering cluster portrait generator, which are connected in sequence; The multi-modal feature extractor is set with a multi-modal feature extraction model. The multi-modal feature extraction model is constructed based on the BERT-ResNet-CNN-RNN-Wav2Vec2.0-AE algorithm. The multi-modal feature extraction model includes a text feature extraction module constructed based on the BERT algorithm, an image feature extraction module constructed based on the ResNet algorithm, a video feature extraction module constructed based on the CNN-RNN-Wav2Vec2.0 algorithm, a user feature extraction module constructed based on the AE algorithm, and a propagation feature extraction module constructed based on the AE algorithm. Among them, the full English name of BERT is "Bidirectional Encoder Representations from Transformers", and the Chinese name is "来自Transformers的双向编码器表示"; the full English name of ResNet is "Residual Network", and the Chinese name is "残差网络"; the full English name of CNN is "Convolutional Neural Network", and the Chinese name is "卷积神经网络"; the full English name of RNN is "Recurrent Neural Network", and the Chinese name is "循环神经网络"; the full English name of Wav2Vec2.0 is "Waveform to Vector ⒉0", and the Chinese name is "波形到向量2.0"; the full English name of AE is "AutoEncoder", and the Chinese name is "自编码器". The spatio-temporal awareness incremental clustering device is set with a spatio-temporal awareness incremental clustering model. The spatio-temporal awareness incremental clustering model is constructed based on the spatio-temporal incremental clustering method (Spatio-temporal Incremental Clustering Method, STICM). The spatio-temporal awareness incremental clustering model is set with a feature clustering module constructed based on the Spatial-Temporal Density-Based Spatial Clustering of Applications with Noise (ST-DBSCAN) connected in sequence, a new data processing module constructed based on the Incremental Processing (IP) algorithm, and a cluster management module constructed based on the Dynamic Cluster Management (DCM) algorithm. The clustering portrait generator is provided with a clustering portrait generation model, which is constructed based on the GCN-DualGAN algorithm, and the clustering portrait generation model includes a clustering relationship construction module constructed based on the GCN algorithm and a clustering portrait generation module constructed based on the DualGAN algorithm, which are connected in sequence. The full name of GCN in English is "Graph Convolutional Network", and its Chinese name is "Graph Convolutional Network", and the full name of DualGAN in English is "DualGenerative Adversarial Network", and its Chinese name is "Dual Generative Adversarial Network". The media data analysis engine is provided with a media data analysis model, which includes a feature selection module based on the Channel Attention (CA) algorithm, a multi-dimensional analysis module based on the Multi-dimensional Analysis (MDA) algorithm, and a report generation module based on the Conditional Generative Adversarial Network (cGAN) algorithm, which are connected in sequence. The multi-dimensional analysis module is provided with several parallel dimensional analysis channels. The dimensional analysis channels include a content topic evolution analysis channel based on the LDA-DTM algorithm, a sentiment distribution analysis channel based on the VADER-LSTM algorithm, a key communication analysis channel based on the SNA algorithm, a user portrait analysis channel based on the GMM-SVM algorithm, a spatiotemporal distribution analysis channel based on the STGNN algorithm, and a media content analysis channel based on the VGG algorithm. LDA stands for "Dynamic Topic Model," DTM stands for "Dual Generative Adversarial Network," VADER stands for "Valence Aware Dictionary and sEntimentReasoner," LSTM stands for "Long Short-Term Memory," SNA stands for "Social Network Analysis," GMM stands for "Gaussian Mixture Model," and SVM stands for "Support Vector Machine", the Chinese name is "Support Vector Machine", the English full name of STGNN is "Spatio-Temporal Graph Neural Network", the Chinese name is "Spatio-Temporal Graph Neural Network", the English full name of VGG is "Visual Geometry Group", the Chinese name is "Visual Geometry Group"; S2: Use the media data crawling engine to generate real-time data crawling strategies for different media platforms, and crawl a number of real-time multimodal media data based on the real-time data crawling strategies, including the following steps: S2-1: Use the media data crawling interface of the media data crawling engine to collect basic information of different media platforms (such as website structure, historical anti-crawling behavior patterns, and data value assessment), and input this basic information into the data crawling strategy generator; S2-2: Based on the basic information, use the data crawling strategy generator of the media data crawling engine to generate a strategy to obtain a real-time data crawling strategy, including the following steps: S2-2-1: Based on the basic information, the platform-level intelligent agent of the platform-level data crawling strategy generation layer of the data crawling strategy generation model is used to generate a strategy in combination with the first multi-optimization target set to obtain a platform-level real-time data crawling strategy. ,in, For the The first, second,..., Platform-level real-time data crawling alternative strategies, is the strategy indicator; S2-2-2: Based on the multi-objective conflict resolution mechanism, the platform status is analyzed according to media data. The coordination agent of the coordination layer of the data crawling strategy generation model is used to resolve the multi-objective conflicts of the platform-level real-time data crawling strategy to obtain the optimized platform-level real-time data crawling strategy, including the following steps: S2-2-2-1: Based on the multi-objective conflict resolution mechanism, the third-generation Non-dominated Sorting Genetic Algorithm III (NSGA-III) algorithm is used to obtain the real-time Pareto frontier solution set from the platform-level real-time data crawling strategy, and obtain the real-time strategy similarity matrix of the real-time Pareto frontier solution set. The steps include: S2-2-2-1-1: Merge all platform-level real-time data crawling alternative strategies involved in the platform-level real-time data crawling strategy into a strategy pool ,in, The total number of platform-level real-time data crawling strategies; S2-2-2-1-2: Using NSGA-III algorithm, from the strategy pool Generate a file containing Pareto frontier solutions of strategies ,in, is the total number of Pareto front solutions; these strategies are Pareto optimal among multiple objectives; S2-2-2-1-3: Calculate the similarity between any two strategies in the Pareto frontier set and construct a similarity matrix , similarity matrix It is symmetrical; Platform-level real-time data crawling strategies may contain multiple objectives, which may conflict with each other. NSGA-III is a multi-objective optimization algorithm that filters Pareto frontier solutions from a given set of platform-level real-time data crawling strategies. Solutions on the Pareto frontier represent "non-dominated" solutions in the current set of strategies that cannot improve any objective without sacrificing at least one other objective. If the strategies in the strategy set of the platform-level real-time data crawling strategy have very different performances on multiple objectives, the Pareto front obtained by NSGA-III may be relatively "scattered", which indirectly reflects that there are obvious multi-objective conflicts in the original strategy set. Conversely, if the Pareto front is relatively "compact", it may indicate that the conflicts are relatively small or the original strategy is already close to Pareto optimality. The Pareto frontier solution set itself also provides a set of potential, multi-objective optimized policy options, which represent different possibilities for trade-offs between conflicting objectives; The similarity matrix quantifies the degree of similarity between different strategies on the Pareto front. Similar strategies may perform similarly on some objectives but differ on others. By analyzing the similarity matrix, we can identify which strategy combinations are likely to cause more serious conflicts (for example, highly similar strategies with significantly different objective values) or which strategies are highly redundant. S2-2-2-2: Integrate the media data analysis platform status, the real-time Pareto frontier solution set, and the real-time strategy similarity matrix to obtain the real-time complete state observation that can be recognized by the coordination agent in the coordination layer of the data crawling strategy generation model ,in, is the encoding representation of the Pareto frontier concentration strategy, For real-time complete state observation, for t The status of the media data analysis platform at all times, is the similarity matrix, t It is the time indication quantity; S2-2-2-3: Map the real-time complete state observation to the coordination state space of the coordination agent to obtain the updated coordination state space of the coordination agent; S2-2-2-4: Based on the updated coordination state space and the preset coordination action space (selecting the strategy in the real-time Pareto frontier solution set), use the coordination agent to resolve multi-objective conflicts, obtain the optimized platform-level real-time data crawling strategy, and transmit the optimized platform-level real-time data crawling strategy to the server-level data crawling strategy generation layer of the data crawling strategy generation model; S2-2-3: Based on the basic information, the status of the media data analysis platform, and the optimized platform-level real-time data crawling strategy, the server-level agent of the server-level data crawling strategy generation layer of the data crawling strategy generation model is used to generate strategies to obtain server-level real-time data crawling strategies for several media platforms. S2-2-4: Integrate the optimized platform-level real-time data crawling strategy and several server-level real-time data crawling strategies to obtain the real-time data crawling strategy for each media platform; S2-3: Based on the real-time data crawling strategy, use the media data crawling executor of the media data crawling engine to crawl a number of real-time multimodal media data of the corresponding media platform; Real-time multimodal media data includes real-time text data, real-time image data, real-time video data, real-time user information data, and real-time interactive data of the media platform; S3: Using a media data processing engine, perform data processing on a number of real-time multimodal media data to obtain a number of real-time media data clusters and their real-time cluster profiles, including the following steps: S3-1: Using a media data preprocessor of a media data processing engine, preprocessing a plurality of real-time multimodal media data from different media platforms to obtain a plurality of preprocessed real-time multimodal media data; Preprocessing involves cleaning and standardizing the collected raw data, including removing noise data (such as HTML tags, special characters, and duplicate data), handling missing values, performing text segmentation, removing stop words, part-of-speech tagging, stemming, and other natural language processing operations. For image and video data, preprocessing includes format conversion and resolution adjustment. Data from different sources and formats are uniformly converted into structured or semi-structured data formats to facilitate subsequent analysis. S3-2: Using a multimodal feature extractor of a media data processing engine, extracting real-time multimodal feature vectors of a plurality of pre-processed real-time multimodal media data, including the following steps: S3-2-1: Inputting the pre-processed real-time multimodal media data into a multimodal feature extraction model of a multimodal feature extractor of a media data processing engine; S3-2-2: Use the text feature extraction module of the multimodal feature extraction model to extract real-time text features from real-time text data in real-time multimodal media data. Real-time text features include semantic vectors that capture deep semantic information, as well as keywords, topic tags, and sentiment polarity. S3-2-3: Use the image feature extraction module of the multimodal feature extraction model to extract real-time image features of real-time image data in real-time multimodal media data; real-time image features are visual feature vectors used to identify image content, scenes, and objects; S3-2-4: using a video feature extraction module of a multimodal feature extraction model to extract real-time video features of real-time video data in real-time multimodal media data; Real-time video features include real-time frame-level features of real-time video data extracted based on the CNN algorithm, real-time time series features of real-time video data extracted based on the RNN algorithm, and real-time audio features of real-time video data extracted based on the Wav2Vec2.0 algorithm. S3-2-6: Use the user feature extraction module of the multimodal feature extraction model to extract real-time user features from real-time user information data in real-time multimodal media data; real-time user features include user ID, number of followers, activity level, and historical content tags; S3-2-7: Use the communication feature extraction module of the multimodal feature extraction model to extract real-time communication features of real-time interactive data in real-time multimodal media data; real-time communication features include the number of forwarding, number of comments, number of likes, number of shares, etc. S3-2-8: Combine the real-time text features, real-time image features, real-time video features, real-time user features, and real-time communication features to obtain the corresponding real-time multimodal feature vector; S3-3: Using the multimodal feature spatiotemporal aligner of the media data processing engine, perform spatiotemporal alignment on a plurality of real-time multimodal feature vectors to obtain a real-time spatiotemporal alignment matrix including the plurality of real-time multimodal feature vectors, including the following steps: S3-3-1: Align the timestamps of all real-time multimodal feature vectors using a unified time base, handle missing or incorrect timestamps (e.g., using interpolation or inference based on content similarity), and obtain several aligned real-time multimodal feature vectors. S3-3-2: Calculate the temporal distance between aligned real-time multimodal feature vectors (taking into account the time decay effect, with more weight given to recent data) and content similarity (based on the vector similarity after multimodal feature fusion, such as using cosine similarity or Siamese network metrics); S3-3-3: Construct a spatiotemporal alignment matrix based on the temporal distance and content similarity of several aligned real-time multimodal feature vectors. Represent the data point of the spatiotemporal alignment matrix corresponding to each aligned real-time multimodal feature vector as an information vector containing its multimodal features, timestamp, spatial location (if available, such as the user's geographic location), and spatiotemporal relationship with other data points. S3-4: Using the spatiotemporal-aware incremental clusterer of the media data processing engine, clustering the real-time multimodal feature vectors in the real-time spatiotemporal alignment matrix to obtain a number of real-time media data clusters; S3-4-1: Use the feature clustering module of the media data processing engine's spatiotemporal-aware incremental clusterer to cluster several real-time multimodal feature vectors in the real-time spatiotemporal alignment matrix within the initial time period, discover spatiotemporal clusters of arbitrary shapes, and obtain several real-time media data clusters. Its core parameters (such as neighborhood radius and minimum number of points) are dynamically adjusted according to data density. S3-4-2: For newly input real-time multimodal feature vectors, incremental updates are performed using the new data processing module; First, the newly input real-time multimodal feature vector is judged for spatiotemporal proximity with the existing clusters (the distance between the new data and the core points of each cluster calculated based on the spatiotemporal alignment matrix); if the newly input real-time multimodal feature vector falls within the neighborhood of a cluster, it is assigned to the cluster, and the spatiotemporal boundaries and features of the cluster are updated; if the newly input real-time multimodal feature vector does not fall into the neighborhood of any cluster, but meets the conditions for forming a new cluster (there are enough new data points around it to form a density connection), a new cluster is created; if the newly input real-time multimodal feature vector neither falls into the existing cluster nor meets the conditions for forming a new cluster, it is temporarily marked as noise or pending data; S3-4-3: Use the cluster management module to manage the spatiotemporal clusters corresponding to several real-time media data clusters; Regularly evaluate the activity of existing clusters (such as the rate of new data inflow and the rate of change of content within the cluster); mark or archive "aging" clusters that have no new data inflow for a long time or have extremely low activity to avoid interference with real-time analysis; monitor cluster evolution, identify dynamic behaviors such as cluster splits, mergers, and migrations, and record their evolutionary trajectories; S3-5: Using the clustering profile generator of the media data processing engine, generate a real-time cluster profile corresponding to each real-time media data cluster, including the following steps: S3-5-1: Define the characteristic dimensions of cluster portraits, including but not limited to: topic information, sentiment information, key figures / institutions, typical user profiles (age, gender, and interest tag inference), dissemination mode (explosive / continuous / cyclical), geographical distribution (if supported by data), spatiotemporal patterns, and relationships with other clusters; S3-5-2: Input each real-time media data cluster into the clustering portrait generation model of the clustering portrait generator of the media data processing engine; S3-5-3: Use the cluster relationship building module of the clustering and cluster portrait generation model to perform relationship modeling on the real-time multimodal feature vectors within the real-time media data clusters, and obtain a set of real-time node embedding vectors including the real-time relationship edges between the real-time multimodal feature vectors; The data points of each real-time multimodal feature vector are regarded as nodes in the graph. The node features are their multimodal fusion features, and the edges represent the similarity or propagation relationship between data points. Through GNN node representation learning, the embedding vector of each data point in low-dimensional space is obtained. This vector can better reflect its structure and semantic information within the cluster. S3-5-4: Based on the portrait feature dimensions, use the cluster portrait generation module of the cluster portrait generation model to generate a portrait of the real-time node embedding vector set to obtain a real-time cluster portrait corresponding to each real-time media data cluster; The clustering portrait generation module consists of a generator G and a discriminator D. The generator G takes as input the real-time node embedding vector set (or cluster aggregate features) learned by the GNN of the cluster, with the goal of generating a complete cluster portrait feature vector. The discriminator D takes as input the real cluster portrait feature vector (labeled by experts or generated by historical data statistics) and the portrait feature vector generated by the generator G, with the goal of distinguishing between true and false portraits. Through adversarial training, the generator G learns to extract the most representative portrait features from the data, while the discriminator D improves its ability to recognize true portraits. Ultimately, the generator can output high-quality, rich, and generalizable cluster portraits, namely cluster portraits. Perform consistency checks and semantic enhancement on the generated real-time cluster profiles to ensure logical rationality between features; associate and store the final real-time cluster profiles (including feature vectors, interpretable labels, confidence levels, etc.) with the corresponding data clusters; S4: Based on a number of real-time media data clusters and their real-time cluster portraits, a media data analysis engine is used to perform multi-dimensional analysis to obtain a real-time media data analysis report; Multi-dimensional analysis includes content theme evolution analysis, sentiment distribution analysis, key communication analysis, user portrait analysis, spatiotemporal distribution analysis, and media content analysis; Based on a number of real-time media data clusters and their real-time cluster portraits, a media data analysis engine is used to perform multi-dimensional analysis to obtain a real-time media data analysis report, including the following steps: S4-1: Input the real-time media data clusters and their real-time cluster profiles into the media data analysis engine, and use the feature selection module of the media data analysis engine to select real-time key features of different dimensions; The key features of the content theme evolution analysis dimension include the text features of each multimodal feature vector in the media data clustering cluster and the theme information of the portrait feature vector in the cluster portrait; the key features of the sentiment distribution analysis dimension include the text features of each multimodal feature vector and the sentiment tendency information of the portrait feature vector in the cluster portrait; the key features of the key communication analysis dimension include the communication features of each multimodal feature vector and the communication mode (explosive / continuous / cyclical) of the portrait feature vector in the cluster portrait; the key features of the user portrait analysis dimension include the user features of each multimodal feature vector and the key figures / institutions of the portrait feature vector in the cluster portrait, typical user portraits (age, gender, interest tag inference) and the relationship with other clusters; the key features of the spatiotemporal distribution analysis dimension include the timestamp, spatial position and spatiotemporal mode of the portrait feature vector in the cluster portrait; the key features of the media content analysis dimension include the image features and video features of each multimodal feature vector; S4-2: Inputting the real-time key features of different dimensions into the corresponding dimensional analysis channel of the multi-dimensional analysis module of the media data analysis engine for analysis to obtain the corresponding real-time dimensional analysis results, including the following steps: S4-2-1: Input the real-time key features of the content theme evolution analysis dimension into the sentiment distribution analysis channel of the multi-dimensional analysis module of the media data analysis engine. Use the LDA algorithm to perform topic modeling based on the real-time text features of each real-time multimodal feature vector, analyze the topic distribution, and combine the topic information in the real-time portrait feature vector in the real-time cluster portrait to analyze and obtain the real-time core topics. Use the DTM algorithm to analyze the change trend of the real-time core topics over time, capture the real-time topic temporal characteristics, and combine the real-time core topics to obtain the real-time content theme evolution analysis results; S4-2-2: Input the real-time key features of the sentiment distribution analysis dimension into the sentiment distribution analysis channel of the multi-dimensional analysis module of the media data analysis engine. Use the VADER algorithm to perform sentiment analysis based on the real-time text features of each real-time multimodal feature vector. Analyze the sentiment classification and combine it with the sentiment tendency information in the real-time portrait feature vector in the real-time cluster portrait to obtain the real-time sentiment classification. Use the LSTM algorithm to analyze the change trend of the real-time sentiment classification over time. Combined with the real-time sentiment classification, obtain the real-time sentiment distribution analysis results. S4-2-3: Input the real-time key features of the key communication analysis dimension into the key communication analysis channel of the multi-dimensional analysis module of the media data analysis engine. Based on the real-time communication features of each real-time multimodal feature vector and the real-time communication mode (burst / continuous / cyclical) of the real-time portrait feature vector in the real-time cluster portrait, construct a real-time communication network of users or content within the cluster. Based on the real-time communication network, use the SNA algorithm to identify key communication nodes and paths, and calculate the efficiency indicators of the real-time communication network (such as average path length and clustering coefficient). This evaluates the communication effect and obtains the real-time key communication analysis results. S4-2-4: Input the real-time key features of the user portrait analysis dimension into the user portrait analysis channel of the multi-dimensional analysis module of the media data analysis engine. Based on the real-time portrait feature vectors in the real-time cluster portrait, the GMM algorithm is used for secondary clustering to identify more fine-grained real-time user groups within the cluster (such as "deep participation", "onlookers", and "opinion leaders"). The SVM is then used to output real-time classification labels (such as user type) for the real-time user groups to obtain the real-time user portrait analysis results. S4-2-5: Input the real-time key features of the spatiotemporal distribution analysis dimension into the spatiotemporal distribution analysis channel of the multi-dimensional analysis module of the media data analysis engine. Based on the timestamp and spatial location of each real-time multimodal feature vector and the spatiotemporal pattern of the real-time portrait feature vector in the real-time cluster portrait, the STGNN algorithm is used to capture the spatiotemporal dependency and perform spatiotemporal distribution analysis to obtain the real-time spatiotemporal distribution analysis results. S4-2-6: Input the real-time key features of the media content analysis dimension into the media content analysis channel of the multi-dimensional analysis module of the media data analysis engine, and use the VGG algorithm to analyze the real-time image features and real-time video features of each real-time multimodal feature vector to obtain real-time media content analysis results; S4-3: Integrate the real-time dimensional analysis results input from all dimensional analysis channels (including real-time content theme evolution analysis results, real-time sentiment distribution analysis results, real-time key communication analysis results, real-time user portrait analysis results, real-time spatiotemporal distribution analysis results, and real-time media content analysis results) to obtain real-time media data clustering and the corresponding real-time cluster portraits. S4-4: Generate a real-time media data analysis report using a report generation module of a media data analysis engine based on the real-time media data analysis results of all real-time media data clusters.

[0020] Example 2: like Figure 2As shown, this embodiment provides an artificial intelligence-based media data analysis system for implementing a media data analysis method. The system is set on a media data analysis platform, and the system includes a media data crawling engine, a media data processing engine, and a media data analysis engine connected in sequence. The media data crawling engine is respectively connected to several external media platforms.

[0021] Media data crawling engine, used to generate real-time data crawling strategies for different media platforms and crawl a number of real-time multimodal media data according to the real-time data crawling strategies; A media data processing engine is used to process a number of real-time multimodal media data to obtain a number of real-time media data clusters and real-time cluster portraits; The media data analysis engine is used to perform multi-dimensional analysis based on a number of real-time media data clusters and their real-time cluster portraits to obtain a real-time media data analysis report.

[0022] The present invention provides an artificial intelligence-based media data analysis method and system, which dynamically generates real-time strategies through an artificial intelligence-driven media data crawling engine and efficiently captures data. Combined with the rapid response of the processing and analysis engine, the entire analysis process can keep up with the real-time dynamics of information dissemination, greatly shortening the time from data generation to analysis result output, solving the core problem of lagging analysis in existing technologies, and being able to instantly capture the rise and evolution of emergencies and hot topics, providing valuable first-time insights for public opinion monitoring, crisis warning, market response, etc. The media data crawling engine can generate strategies for different media platforms, effectively covering mainstream and emerging platforms, avoiding information blind spots, making the analysis results more representative, and can dynamically adjust the crawling strategy according to data feedback and goals, automatically adapting to platform changes and hot spot shifts, improving the efficiency and pertinence of data acquisition, and enhancing data coverage. comprehensiveness and intelligence; the media data processing engine can uniformly process multiple modal data such as text, images, and videos, and identify the relationships between them, providing a more three-dimensional and complete information view, and can effectively extract and analyze the spatiotemporal attributes of data, revealing the geographical distribution and temporal patterns of information dissemination, and providing a new dimension for understanding the dissemination mechanism. The feature selection, multi-dimensional analysis and other modules of the media data processing engine are all driven by artificial intelligence, which can automatically identify key information, conduct more in-depth and objective analysis, and reduce human intervention and subjective bias; the media data analysis engine conducts media data analysis from multiple dimensions such as content theme evolution analysis, sentiment distribution analysis, key communication analysis, user portrait analysis, spatiotemporal distribution analysis, and media content analysis, which improves the value of media data analysis results and obtains a systematic integrated analysis report, improving the convenience and efficiency of media data analysis.

[0023] The present invention is not limited to the above optional embodiments. Anyone can derive various other forms of products based on the teachings of the present invention. The above specific embodiments should not be construed as limiting the scope of protection of the present invention. The scope of protection of the present invention shall be based on the scope defined in the claims, and the description can be used to interpret the claims.

Claims

1. A media data analysis method based on artificial intelligence, characterized by: The steps include: Based on artificial intelligence algorithms, build media data crawling engines, media data processing engines, and media data analysis engines in the media data analysis platform; Use the media data crawling engine to generate real-time data crawling strategies for different media platforms, and crawl a number of real-time multimodal media data based on the real-time data crawling strategies; Using a media data processing engine, processing a number of real-time multimodal media data to obtain a number of real-time media data clusters and real-time cluster portraits; Based on a number of real-time media data clusters and their real-time cluster portraits, a media data analysis engine is used to perform multi-dimensional analysis to obtain a real-time media data analysis report.

2. The method for analyzing media data based on artificial intelligence according to claim 1, characterized in that: The real-time multimodal media data includes real-time text data, real-time image data, real-time video data, real-time user information data and real-time interactive data of the media platform.

3. The method for analyzing media data based on artificial intelligence according to claim 2, characterized in that: The media data crawling engine is provided with a media data crawling interface, a media data crawling executor, and a data crawling strategy generator connected in sequence. The media data crawling interface is respectively connected to several external media platforms. The data crawling strategy generator is provided with a data crawling strategy generation model. The data crawling strategy generation model is constructed based on the MAPGRPO algorithm, and the data crawling strategy generation model includes a platform-level data crawling strategy generation layer, a coordination layer and a server-level data crawling strategy generation layer connected in sequence. The platform-level data crawling strategy generation layer is provided with a first multi-optimization target set and a platform-level intelligent agent, the coordination layer is provided with a coordination intelligent agent and a multi-target conflict resolution mechanism, and the server-level data crawling strategy generation layer is provided with a second multi-optimization target set and several parallel server-level intelligent agents.

4. The method for analyzing media data based on artificial intelligence according to claim 3, characterized in that: The media data processing engine is provided with a media data preprocessor, a multimodal feature extractor, a multimodal feature spatiotemporal aligner, a spatiotemporal perception incremental clusterer and a clustering cluster portrait generator connected in sequence; The multimodal feature extractor is provided with a multimodal feature extraction model, which is constructed based on the BERT-ResNet-CNN-RNN-Wav2Vec2.0-AE algorithm, and the multimodal feature extraction model includes a text feature extraction module constructed based on the BERT algorithm, an image feature extraction module constructed based on the ResNet algorithm, a video feature extraction module constructed based on the CNN-RNN-Wav2Vec2.0 algorithm, a user feature extraction module constructed based on the AE algorithm, and a propagation feature extraction module constructed based on the AE algorithm; The spatiotemporal-aware incremental clusterer is provided with a spatiotemporal-aware incremental clustering model, which is constructed based on the STICM algorithm, and the spatiotemporal-aware incremental clustering model is provided with a feature clustering module constructed based on the ST-DBSCAN algorithm, a new data processing module constructed based on the IP algorithm, and a cluster management module constructed based on the DCM algorithm. The cluster portrait generator is provided with a cluster portrait generation model, which is constructed based on the GCN-DualGAN algorithm, and the cluster portrait generation model includes a cluster relationship construction module constructed based on the GCN algorithm and a cluster portrait generation module constructed based on the DualGAN algorithm, which are connected in sequence.

5. The method for analyzing media data based on artificial intelligence according to claim 4, characterized in that: The media data analysis engine is provided with a media data analysis model, which includes a feature selection module based on the CA algorithm, a multi-dimensional analysis module based on the MDA algorithm, and a report generation module based on the cGAN algorithm, which are connected in sequence. The multi-dimensional analysis module is provided with several parallel dimensional analysis channels. The dimensional analysis channels include a content topic evolution analysis channel constructed based on the LDA-DTM algorithm, a sentiment distribution analysis channel constructed based on the VADER-LSTM algorithm, a key communication analysis channel constructed based on the SNA algorithm, a user portrait analysis channel constructed based on the GMM-SVM algorithm, a spatiotemporal distribution analysis channel constructed based on the STGNN algorithm, and a media content analysis channel constructed based on the VGG algorithm.

6. The method for analyzing media data based on artificial intelligence according to claim 5, characterized in that: Use a media data crawling engine to generate real-time data crawling strategies for different media platforms. Then crawl a number of real-time multimodal media data based on the real-time data crawling strategies, including the following steps: Use the media data crawling interface of the media data crawling engine to collect basic information from different media platforms and input the basic information into the data crawling strategy generator; Based on the basic information, use the data crawling strategy generator of the media data crawling engine to generate strategies and obtain real-time data crawling strategies; According to the real-time data crawling strategy, the media data crawling executor of the media data crawling engine is used to crawl a number of real-time multimodal media data of the corresponding media platform.

7. The method for analyzing media data based on artificial intelligence according to claim 6, characterized in that: Using a media data processing engine, a plurality of real-time multimodal media data are processed to obtain a plurality of real-time media data clusters and real-time cluster portraits, including the following steps: Using a media data preprocessor of a media data processing engine, preprocessing a plurality of real-time multimodal media data from different media platforms to obtain a plurality of preprocessed real-time multimodal media data; Extracting real-time multimodal feature vectors of a plurality of pre-processed real-time multimodal media data using a multimodal feature extractor of a media data processing engine; Using a multimodal feature spatiotemporal aligner of a media data processing engine, performing spatiotemporal alignment on a plurality of real-time multimodal feature vectors to obtain a real-time spatiotemporal alignment matrix including the plurality of real-time multimodal feature vectors; Using the time-space-aware incremental clusterer of the media data processing engine, clustering a number of real-time multimodal feature vectors in the real-time time-space alignment matrix to obtain a number of real-time media data clusters; The clustering portrait generator of the media data processing engine is used to generate a real-time cluster portrait corresponding to each real-time media data cluster.

8. The method for analyzing media data based on artificial intelligence according to claim 7, characterized in that: The multi-dimensional analysis includes content theme evolution analysis, sentiment distribution analysis, key communication analysis, user portrait analysis, spatiotemporal distribution analysis, and media content analysis.

9. The method for analyzing media data based on artificial intelligence according to claim 8, characterized in that: Based on a number of real-time media data clusters and their real-time cluster portraits, a media data analysis engine is used to perform multi-dimensional analysis to obtain a real-time media data analysis report, including the following steps: Input the real-time media data clusters and their real-time cluster portraits into the media data analysis engine, and use the feature selection module of the media data analysis engine to select real-time key features of different dimensions; Input the real-time key features of different dimensions into the corresponding dimension analysis channel of the multi-dimensional analysis module of the media data analysis engine for analysis to obtain the corresponding real-time dimension analysis results; Integrate the real-time dimensional analysis results input by all dimensional analysis channels to obtain the real-time media data clustering clusters and their real-time cluster portraits corresponding to the real-time media data analysis results; A real-time media data analysis report is generated using a report generation module of a media data analysis engine according to the real-time media data analysis results of all real-time media data clusters.

10. An artificial intelligence-based media data analysis system, for implementing the media data analysis method according to any one of claims 1 to 9, characterized in that: The system is set up on a media data analysis platform, and the system includes a media data crawling engine, a media data processing engine and a media data analysis engine connected in sequence. The media data crawling engine is respectively connected to several external media platforms.

Images