Big data prediction method and platform based on public opinions, computer equipment and storage medium

Abstract

The embodiment of the invention relates to the technical field of big data processing, and discloses a public opinion-based big data prediction method, which comprises the following steps of: acquiring initial financial public opinion data and traditional financial data in real time through a multivariate heterogeneous data acquisition layer; performing multi-language nested sentiment analysis on the initial financial public opinion data according to a dynamic sentiment quantization engine to obtain a text sentiment analysis result, a picture sentiment analysis result and an audio and video sentiment analysis result; performing cross-modal emotion resonance detection and emotion fusion according to a text emotion analysis result, a picture emotion analysis result and an audio and video emotion analysis result; calculating the information propagation speed and range of the initial financial public opinion data according to a space-time propagation network model; and inputting the current currency exchange rate information, the current market emotion index, the information propagation speed and range and the traditional financial data into an exchange rate change prediction model to obtain the exchange rate change of the target currency in a future preset time period. According to the embodiment of the invention, real-time accurate prediction of the exchange rate is realized.

Description

Technical Field

[0001] Embodiments of the present invention relate to the technical field of big data processing, and particularly to a big data prediction method based on public opinion, a big data prediction platform based on public opinion, a computer device, and a computer storage medium. Background Art

[0002] With the development of the Internet, more and more people begin to pay attention to the financial market and express their views on the financial market, resulting in an increasing amount of information in the financial field, which is also known as financial public opinion. Traditional exchange rate prediction methods mainly rely on the experience of experts for judgment, but this method is greatly affected by personal experience. Moreover, with the increase in the amount of information, true and false data are mixed, making it difficult to capture the laws hidden in a large amount of financial public opinion information. Therefore, how to use financial public opinion information to assist in predicting exchange rates has become an important research direction. Summary of the Invention

[0003] In view of the above problems, embodiments of the present invention provide a big data prediction method based on public opinion, a big data prediction platform based on public opinion, a computer device, and a computer storage medium, which are used to solve the problem of low accuracy of exchange rate prediction relying on expert experience in the prior art.

[0004] According to one aspect of the embodiments of the present invention, a big data prediction method based on public opinion is provided, and the method includes:

[0005] Collecting initial financial public opinion data and traditional financial data in real time through a multi-source heterogeneous data collection layer; the public opinion data includes text data, picture data, and audio-video data of multiple social media, news websites, forums, policy text data, and video platforms;

[0006] Performing multi-language nested sentiment analysis on the initial financial public opinion data according to a dynamic sentiment quantification engine to obtain text sentiment analysis results, picture sentiment analysis results, and audio-video sentiment analysis results;

[0007] Performing cross-modal sentiment resonance detection and sentiment fusion according to the text sentiment analysis results, the picture sentiment analysis results, and the audio-video sentiment analysis results to obtain a current market sentiment index; the current market sentiment index is an optimism value or a pessimism value;

[0008] Calculating the information propagation speed and range of the initial financial public opinion data according to a spatio-temporal propagation network model;

[0009] Inputting the current currency exchange rate information, the current market sentiment index, the information propagation speed and range, and the traditional financial data into an exchange rate change prediction model to obtain the exchange rate change of the target currency within a preset future time period.

[0010] In an alternative approach, the multi - language nested sentiment analysis of the initial financial public opinion data according to the dynamic sentiment quantification engine to obtain text sentiment analysis results, picture sentiment analysis results, and audio - video sentiment analysis results includes:

[0011] Obtaining the explicit sentiment corresponding to the text data through the BERT model and the explicit sentiment recognition model; the text sentiment recognition model is trained based on explicit text samples and explicit sentiment labels for the LSTM model;

[0012] Using the TransR model to calculate the semantic distance between the text data in the initial financial public opinion data and the metaphor knowledge graph to determine the metaphor sentiment of the text data; the metaphor knowledge graph includes multiple nodes, each node being a financial metaphor triple, and the elements of the financial metaphor triple being metaphor text, corresponding explicit meaning, and relationship;

[0013] Analyzing the text data in the initial financial public opinion data according to a preset dynamic dictionary and attention weight analysis algorithm to obtain policy expectation game information; the preset dynamic dictionary is constructed in advance based on policy vocabulary;

[0014] Determining the text sentiment analysis result based on the explicit sentiment, metaphor sentiment signal, and policy expectation game sentiment information.

[0015] In an alternative approach, the obtaining of the explicit sentiment corresponding to the text data through the BERT model and the explicit sentiment recognition model further includes:

[0016] Encoding the text data in the initial financial public opinion data through the BERT model to extract context embedding vectors;

[0017] Inputting the context embedding vectors into the explicit sentiment recognition model to obtain the explicit sentiment corresponding to the target currency in the text data; the text sentiment recognition model is trained based on explicit text samples and explicit sentiment labels for the LSTM model.

[0018] In an alternative approach, the method of using the TransR model to calculate the semantic distance between the text data in the initial financial public opinion data and the metaphor knowledge graph to determine the metaphor sentiment of the text data includes:

[0019] Matching the text data in the initial financial public opinion data with the metaphor triples of the metaphor knowledge graph to obtain the matching text in the text data;

[0020] Calculate the semantic distance between the matching text and the metaphor knowledge graph using the TransR model; wherein, the TransR model is iteratively trained according to a preset text sample according to a preset loss function; the preset loss function is:

[0021]

[0022] where S metaphor is the loss function; h r = M r ·h is the mapping of the metaphor text h under the relation r; t r = M r ·t is the mapping of the literal meaning t under the relation r; h is the metaphor text in the metaphor triple, r is the relation in the metaphor triple, and t is the literal meaning in the metaphor triple;

[0023] When the semantic distance is less than the preset distance, obtain the literal meaning corresponding to the matching text;

[0024] Determine the metaphor emotion corresponding to the text data according to the literal meaning.

[0025] In an alternative manner, the cross-modal emotion resonance detection based on the text emotion analysis result, the picture emotion analysis result, and the audio-visual emotion analysis result to obtain the current market emotion indicator includes:

[0026] Align the text emotion analysis result, the picture emotion analysis result, and the audio-visual emotion analysis result to the same feature space;

[0027] Calculate the similarity between the text emotion analysis result and the picture emotion analysis result and the audio-visual emotion analysis result respectively. When there is an emotion conflict, determine the authenticity of the initial financial public opinion data;

[0028] When there is no emotion conflict, perform weighted fusion calculation on the text emotion analysis result, the picture emotion analysis result, and the audio-visual emotion analysis result to obtain the current market emotion indicator.

[0029] In an alternative manner, before the multi-language nested emotion analysis of the initial financial public opinion data by the dynamic emotion quantification engine to obtain the text emotion analysis result, the picture emotion analysis result, and the audio-visual emotion analysis result, the method includes:

[0030] Extract the text content corresponding to the audio-visual data in the initial financial public opinion data;

[0031] Extract the background music rhythm corresponding to the audio-visual data in the initial financial public opinion data;

[0032] Determine the audio-visual emotion analysis result according to the text content and the background music rhythm.

[0033] In an alternative manner, the determining the audio-visual emotion analysis result according to the text content and the background music rhythm includes:

[0034] Encode the text content through a BERT model and input it into an audio-visual emotion recognition model to obtain the first audio-visual emotion corresponding to the text content;

[0035] Input the background music rhythm into a panic correlation model to obtain a panic correlation index; wherein, the panic correlation model is obtained by pre-modeling the correlation between the historical short video background music rhythm and the market panic index;

[0036] Adjust the weight of the first audio-visual emotion according to the panic correlation index to obtain the audio-visual emotion analysis result.

[0037] According to another aspect of the embodiments of the present invention, a big data prediction platform based on public opinion is provided, including:

[0038] A collection module for collecting initial financial public opinion data and traditional financial data in real time through a multi-source heterogeneous data collection layer; the public opinion data includes text data, picture data, and audio-visual data of multiple social media, news websites, forums, policy text data, and video platforms;

[0039] An emotion analysis module for performing multi-language nested emotion analysis on the initial financial public opinion data according to a dynamic sentiment quantification engine to obtain text emotion analysis results, picture emotion analysis results, and audio-visual emotion analysis results;

[0040] An index calculation module for performing cross-modal emotion resonance detection and emotion fusion according to the text emotion analysis result, the picture emotion analysis result, and the audio-visual emotion analysis result to obtain the current market emotion index; the current market emotion index is an optimism value or a pessimism value;

[0041] A propagation calculation module for calculating the information propagation speed and range of the initial financial public opinion data according to a spatio-temporal propagation network model;

[0042] A prediction module for inputting the current currency exchange rate information, the current market emotion index, the information propagation speed and range, and the traditional financial data into an exchange rate change prediction model to obtain the exchange rate change of the target currency within a preset future time period.

[0043] According to another aspect of the embodiments of the present invention, there is provided a computer device, including: a processor, a memory, a communication interface, and a communication bus, and the processor, the memory, and the communication interface complete communication with each other through the communication bus;

[0044] The memory is used to store at least one executable instruction, and the executable instruction causes the processor to execute the operations of the big data prediction method based on public opinion.

[0045] According to still another aspect of the embodiments of the present invention, there is provided a computer-readable storage medium, in which at least one executable instruction is stored, and when the executable instruction runs on a computer device, the computer device is caused to execute the operations of the big data prediction method based on public opinion.

[0046] The above technical solutions provided by the embodiments of the present application have the following advantages compared with the prior art: The initial financial public opinion data and traditional financial data are collected in real time by the multi-source heterogeneous data collection layer in the embodiments of the present application; the initial financial public opinion data is subjected to multi-language nested sentiment analysis by the dynamic sentiment quantification engine to obtain a text sentiment analysis result, a picture sentiment analysis result, and an audio-video sentiment analysis result; cross-modal sentiment resonance detection and sentiment fusion are performed according to the text sentiment analysis result, the picture sentiment analysis result, and the audio-video sentiment analysis result to obtain a current market sentiment index; the current market sentiment index is an optimism value or a pessimism value; the information propagation speed and range of the initial financial public opinion data are calculated according to the spatio-temporal propagation network model; the exchange rate change of the target currency within a preset future time period is obtained by inputting the current currency exchange rate information, the current market sentiment index, the information propagation speed and range, and the traditional financial data into the exchange rate change prediction model. It can improve the efficiency and accuracy of financial public opinion data analysis, thereby improving the accuracy and reliability of exchange rate prediction.

[0047] In addition, the embodiments of the present application also combine the influence of the background music of the audio-video on public opinion, and use a panic correlation model to predict the panic correlation index, so as to predict the influence of public opinion on people's emotions and fuse it into the first audio-video emotion result, making the analysis of the audio-video emotion result more accurate.

[0048] The above description is only an overview of the technical solutions of the embodiments of the present invention. In order to be able to understand the technical means of the embodiments of the present invention more clearly, it can be implemented according to the content of the description. And in order to make the above and other purposes, features, and advantages of the embodiments of the present invention more obvious and understandable, the specific embodiments of the present invention are hereinafter specifically exemplified. Description of the Drawings

[0049] The accompanying drawings are only used to illustrate the embodiments and are not considered as limiting the present invention. Moreover, throughout the drawings, the same reference numerals are used to represent the same components. In the drawings:

[0050] Figure 1 It shows a schematic flowchart of the big data prediction method based on public opinion provided by an embodiment of the present invention;

[0051] Figure 2 It shows a schematic structural diagram of the big data prediction platform based on public opinion provided by an embodiment of the present invention;

[0052] Figure 3 It shows a schematic structural diagram of the computer device provided by an embodiment of the present invention. Detailed Embodiments

[0053] The exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein.

[0054] See Figure 1 As shown, an embodiment of the present application provides a big data prediction method based on public opinion, and the method includes the following steps:

[0055] S110: Collect the initial financial public opinion data and traditional financial data collected in real time through the multi-source heterogeneous data collection layer.

[0056] Specifically, an embodiment of the present invention can collect the initial financial public opinion data and traditional financial data collected in real time through the multi-source heterogeneous data collection layer. Among them, the initial financial public opinion data includes, but is not limited to, text data, picture data, audio and video data, etc. of multiple social media, news websites, forums, policy text data, and short video platforms. The traditional financial data includes, but is not limited to, currency exchange rate information, stock price information, bond price information, futures price information, fund net value information, foreign exchange reserve information, GDP growth rate information, exchange rate information, etc. Among them, the financial data in these information will affect people's expectations for various currencies. Therefore, the sentiment indicators in the phased financial public opinion data are very important.

[0057] Among them, the multi-source heterogeneous data collection layer obtains data from various data sources, including real-time data and offline data, and performs preliminary processing on the collected data, such as format conversion, filtering, deduplication, etc. Finally, the processed data is transmitted to the data storage layer or the data processing layer.

[0058] S120: Perform multi-language nested sentiment analysis on the initial financial public opinion data according to the dynamic sentiment quantification engine to obtain text sentiment analysis results, picture sentiment analysis results, and audio and video sentiment analysis results.

[0059] Among them, the embodiments of the present invention can perform multi-language nested sentiment analysis on the initial financial public opinion data according to a dynamic sentiment quantification engine to obtain text sentiment analysis results, picture sentiment analysis results, and audio-visual sentiment analysis results. Among them, the dynamic sentiment quantification engine includes, but is not limited to, an explicit emotion recognition model, a metaphor emotion recognition model, a policy expectation game emotion recognition model, an audio-visual emotion recognition model, etc. For example, the explicit emotion recognition model is obtained by pre-training a Long Short-Term Memory (LSTM) model according to explicit text samples and explicit emotion labels, the metaphor emotion recognition model is obtained by pre-training a TransR model according to metaphor text samples and explicit meaning labels, and the policy expectation game emotion recognition model is obtained by pre-training an Attention Weighted Analysis (AWA) algorithm according to policy text samples and policy expectation game labels. The audio-visual emotion recognition model is obtained by pre-training a BERT (Bidirectional Encoder Representations from Transformers) model according to audio-visual samples and audio-visual emotion labels. The picture emotion recognition model is obtained by pre-training a BERT model according to picture samples and picture emotion labels.

[0060] Specifically, the embodiments of the present invention obtain text sentiment analysis results in the following manner;

[0061] S1201: Obtain the explicit emotion corresponding to the text data through the BERT model and the explicit emotion recognition model, where the text emotion recognition model is obtained by pre-training an LSTM model according to explicit text samples and explicit emotion labels. For example, the embodiments of the present invention can encode the text data in the initial financial public opinion data through the BERT model to extract context embedding vectors; then input the context embedding vectors into the explicit emotion recognition model to obtain the explicit emotion corresponding to the target currency in the text data. Among them, the explicit text is the text whose meaning can be directly seen, such as the direct expression of "the yen has plummeted"; the explicit emotion refers to the emotion corresponding to the explicit text, including at least optimistic emotion, pessimistic emotion, and neutral emotion.

[0062] S1202: Use the TransR model to calculate the semantic distance between the text data in the initial financial public opinion data and the metaphor knowledge graph, and determine the metaphor emotion of the text data.

[0063] Specifically, in the embodiments of the present invention, a metaphor knowledge graph is pre-constructed. The metaphor knowledge graph includes multiple nodes, and each node is a financial metaphor triple. The elements of the financial metaphor triple are metaphor texts, corresponding explicit meanings, and relationships. For example, the crisis implication of "black swan event" can be expressed as: (black swan, represents, unpredictable risk); or "quantitative easing cliff" can be expressed as: (quantitative easing cliff, associated with, soaring treasury yields).

[0064] Among them, the text data in the initial financial public opinion data is matched with the metaphor triples in the metaphor knowledge graph to obtain the matching text in the text data; then the TransR model is used to calculate the semantic distance between the matching text and the metaphor knowledge graph. Among them, the TransR model is iteratively trained according to a preset text sample according to a preset loss function.

[0065] In an embodiment of the present invention, the preset loss function can be:

[0066]

[0067] Among them, S metaphor is the loss function; h r = M r ·h is the mapping of the metaphor text h under the relationship r; t r = M r ·t is the mapping of the explicit meaning t under the relationship r; h is the metaphor text in the metaphor triple, r is the relationship in the metaphor triple, and t is the explicit meaning in the metaphor triple.

[0068] When the semantic distance is less than the preset distance, the explicit meaning corresponding to the matching text is obtained; then the metaphor emotion corresponding to the text data is determined according to the explicit meaning.

[0069] In this way, public opinion information can be accurately mined, and the accuracy rate of public opinion can be provided. For example, when detecting "the Bank of Japan is walking a tightrope": matching "walking a tightrope → policy dilemma" in the knowledge graph can trigger a 15% increase in the yen policy uncertainty index.

[0070] S1204: Analyze the text data in the initial financial public opinion data according to the preset dynamic dictionary and attention weight analysis algorithm to obtain policy expectation game information.

[0071] Among them, the preset dynamic dictionary is pre-constructed according to policy vocabulary.

[0072] For example, first, a dynamic dictionary is pre-constructed, which can be expressed as:

[0073] Hawkish terms Dovish terms Weight Decisive action Maintain patience 0.9 Intolerable inflation Temporary factors 1.2 Front-loaded interest rate hikes Data-dependent 0.8

[0074] Use Hierarchical Attention Networks to analyze attention weights, identify keyword intensities through word-level attention, locate the core policy statement paragraphs through sentence-level attention, output the "Fed Hawk-Dove Index" (0-100), and obtain policy expectation game information.

[0075] S1205: Determine the text emotion analysis result according to the explicit emotion, metaphorical emotion signal, and policy expectation game emotion information.

[0076] Among them, when analyzing the emotions of audio and video, it is also achieved through the following methods:

[0077] Extract the text content corresponding to the audio and video data in the initial financial public opinion data; extract the background music rhythm corresponding to the audio and video data in the initial financial public opinion data; determine the audio and video emotion analysis result according to the text content and the background music rhythm.

[0078] Specifically, encode the text content through the BERT model and input it into the audio and video emotion recognition model to obtain the first audio and video emotion corresponding to the text content; input the background music rhythm into the panic correlation model to obtain the panic correlation index; among them, the panic correlation model is pre-modeled according to the historical short video background music rhythm and the market panic index; adjust the weight of the first audio and video emotion according to the panic correlation index to obtain the audio and video emotion analysis result. In the embodiment of the present invention, the panic correlation model is used to analyze the relationship between the background music rhythm and the market panic index, so as to provide a basis for weight adjustment of the emotion analysis result. Among them, a regression model can be used to construct the panic correlation model in an embodiment of the present invention. If the panic correlation index is higher than the preset panic threshold, it indicates that the background music rhythm has a strong correlation with the market panic emotion, so the weight of the audio and video emotion analysis result is appropriately reduced. If the panic correlation index is low, the weight of the audio and video emotion analysis result remains unchanged.

[0079] S130: Perform cross-modal emotion resonance detection and emotion fusion according to the text emotion analysis result, the picture emotion analysis result, and the audio and video emotion analysis result to obtain the current market emotion index; the current market emotion index is an optimism value or a pessimism value.

[0080] Among them, cross-modal emotion resonance detection and emotion fusion are performed through the following specific methods:

[0081] First, the text emotion analysis results, the image emotion analysis results, and the audio and video emotion analysis results are aligned to the same feature space. The alignment can be text-centered: the image and audio and video features are projected into the text feature space through the MRAN model (Multimodal Reconstruction and Alignment Network), and the alignment is performed by minimizing the distance between the features and the emotion category embedding.

[0082] Then, the similarity between the text emotion analysis result and the picture emotion analysis result and the audio and video emotion analysis result is calculated respectively. When there is an emotion conflict, the authenticity of the initial financial public opinion data is determined. Specifically, in one embodiment of the present invention, the similarity between the text emotion analysis result and the picture emotion analysis result, and the similarity between the text emotion analysis result and the audio and video emotion analysis result can be calculated by cosine similarity. The cosine similarity can be expressed as:

[0083]

[0084] in, and Respectively represent the emotion feature vectors of the two modes. For example, it can be the text emotion feature vector in the text emotion analysis result and the image emotion feature vector in the image emotion analysis result.

[0085] In another embodiment of the present invention, a multimodal deep learning model such as the CLIP model (Contrastive Language-Image Pre-training) can also be used to calculate the similarity between the text emotion analysis results, the picture emotion analysis results, and the audio and video emotion analysis results.

[0086] Among them, when there are conflicts in the sentiment analysis results of different modalities, it is necessary to judge the authenticity of the financial public opinion data, eliminate inconsistent data, and conduct sentiment analysis. Specifically, this is achieved through the following methods: data source verification, inter-modal relationship analysis, external verification, dynamic weight adjustment, and conflict resolution strategy.

[0087] Among them, for data source verification: check whether the news source in the text data is an authoritative media (such as Reuters, Bloomberg, Xinhua News Agency, etc.), and whether the social media data comes from trustworthy users; multi-source data comparison: obtain information about the same event from multiple channels and compare the sentiment analysis results from different sources. For image and video data, verify the origin of the image or video and whether it has been tampered with or forged. For audio data, confirm whether the audio source is reliable and whether it is consistent with the text content. Authoritative information reference: introduce official data, industry reports or expert opinions for verification. Historical data comparison: combine historical public opinion data to determine whether the current conflict conforms to historical trends. For public opinion data with high risks or high conflicts, introduce an artificial review mechanism and combine expert opinions for the final judgment.

[0088] In the embodiments of the present invention, when two of the emotions are consistent and the remaining one is in conflict, it is possible to choose not to eliminate the data but modify the weight value during the corresponding weighted fusion. If the text and image emotions are consistent but the audio emotion is in conflict, then reduce the weight of the audio.

[0089] Among them, when there is no emotional conflict, perform a weighted fusion calculation on the text emotion analysis result, the picture emotion analysis result, and the audio-visual emotion analysis result to obtain the current market sentiment indicator. Specifically, in the financial field, text data (such as news reports) is usually more reliable than images or audio, so a higher weight can be assigned.

[0090] Among them, after fusion, the current market sentiment indicator is obtained, and this indicator can be an optimism value or a pessimism value.

[0091] S140: Calculate the information propagation speed and range of the initial financial public opinion data according to the spatio-temporal propagation network model.

[0092] The spatio-temporal propagation network model is established in advance based on historical financial public opinion data. Specifically, divide the financial public opinion data samples into multiple time windows and model the data within each time window. The hyperedges within each slice represent the propagation relationship of information between nodes (such as users, topics, etc.). Stack all the hypernetwork slices to form a complete hypernetwork. By analyzing the structure and dynamic changes of the hypernetwork, capture the spatio-temporal characteristics of information propagation. Among them, within each time window, identify the propagation path and node relationship of the information. Use the shortest path algorithm (such as Dijkstra's algorithm) to calculate the shortest path of the information from the source node to the target node and record the propagation time of each path. Among them, the propagation speed = path length / propagation time. The path length is the distance between nodes. Calculate the information propagation speed by analyzing the propagation path and time delay of the information in the hypernetwork.

[0093] After that, by analyzing the closeness centrality of nodes in the hypernetwork, key nodes and paths of information dissemination are identified. The dissemination range is evaluated by the connectivity of nodes, that is, the number of nodes that information can reach.

[0094] S150: Input the current currency exchange rate information, the current market sentiment indicator, the information dissemination speed and range, and the traditional financial data into an exchange rate change prediction model to obtain the exchange rate change of the target currency within a preset future time period.

[0095] Specifically, this application can input the current currency exchange rate information, the current market sentiment indicator, the information dissemination speed and range, and the traditional financial data into an exchange rate change prediction model to obtain the exchange rate change of the target currency within a preset future time period. Among them, the exchange rate change prediction model is pre-trained based on historical exchange rate data, historical market sentiment indicators, historical information dissemination speed and range, and historical financial public opinion data. Among them, the model is trained by combining ARMA-GARCH and Markov chains because the ARMA-GARCH model can be used to capture the dynamic characteristics of exchange rate fluctuations. The Markov chain can also be used to predict the expected value and interval of the exchange rate.

[0096] An embodiment of this application discloses a big data prediction method based on public opinion, and the method includes: real-time collecting initial financial public opinion data and traditional financial data through a multi-source heterogeneous data collection layer; performing multi-language nested sentiment analysis on the initial financial public opinion data according to a dynamic sentiment quantification engine to obtain text sentiment analysis results, picture sentiment analysis results, and audio-visual sentiment analysis results; performing cross-modal sentiment resonance detection and sentiment fusion according to the text sentiment analysis results, the picture sentiment analysis results, and the audio-visual sentiment analysis results to obtain a current market sentiment indicator; the current market sentiment indicator is an optimism value or a pessimism value; calculating the information dissemination speed and range of the initial financial public opinion data according to a spatio-temporal dissemination network model; inputting the current currency exchange rate information, the current market sentiment indicator, the information dissemination speed and range, and the traditional financial data into an exchange rate change prediction model to obtain the exchange rate change of the target currency within a preset future time period. This application can improve the efficiency and accuracy of financial public opinion data analysis, thereby improving the accuracy and reliability of financial prediction.

[0097] The above technical solution provided by the embodiments of the present application has the following advantages compared with the prior art: The embodiments of the present application collect initial financial public opinion data and traditional financial data in real time through a multi-source heterogeneous data collection layer; perform multi-language nested sentiment analysis on the initial financial public opinion data according to a dynamic sentiment quantification engine to obtain text sentiment analysis results, picture sentiment analysis results, and audio-video sentiment analysis results; perform cross-modal sentiment resonance detection and sentiment fusion according to the text sentiment analysis results, the picture sentiment analysis results, and the audio-video sentiment analysis results to obtain the current market sentiment index; the current market sentiment index is an optimism value or a pessimism value; calculate the information propagation speed and range of the initial financial public opinion data according to a spatio-temporal propagation network model; input the current currency exchange rate information, the current market sentiment index, the information propagation speed and range, and the traditional financial data into an exchange rate change prediction model to obtain the exchange rate change of the target currency within a preset future time period. It can improve the efficiency and accuracy of financial public opinion data analysis, thereby improving the accuracy and reliability of exchange rate prediction.

[0098] In addition, the embodiments of the present application also consider the impact of the background music of audio-visual content on public opinion. By using a panic correlation model, the panic correlation index is predicted, so as to predict the impact of public opinion on people's emotions and integrate it into the first audio-video emotion result, making the analysis of the audio-video emotion result more accurate.

[0099] Figure 2 FIG. shows a schematic structural diagram of a big data prediction platform based on public opinion provided by an embodiment of the present invention. As Figure 2 shown, the platform 200 includes:

[0100] A collection module 210, configured to collect initial financial public opinion data and traditional financial data in real time through a multi-source heterogeneous data collection layer; the public opinion data includes text data, picture data, and audio-video data of multiple social media, news websites, forums, policy text data, and video platforms;

[0101] An emotion analysis module 220, configured to perform multi-language nested sentiment analysis on the initial financial public opinion data according to a dynamic sentiment quantification engine to obtain text sentiment analysis results, picture sentiment analysis results, and audio-video sentiment analysis results;

[0102] An index calculation module 230, configured to perform cross-modal sentiment resonance detection and sentiment fusion according to the text sentiment analysis results, the picture sentiment analysis results, and the audio-video sentiment analysis results to obtain the current market sentiment index; the current market sentiment index is an optimism value or a pessimism value;

[0103] A propagation calculation module 240, configured to calculate the information propagation speed and range of the initial financial public opinion data according to a spatio-temporal propagation network model;

[0104] A prediction module 250 is configured to input the current currency exchange rate information, the current market sentiment indicator, the information dissemination speed and scope, and the traditional financial data into an exchange rate change prediction model to obtain the exchange rate change of the target currency within a preset future time period.

[0105] The above technical solution provided by the embodiments of the present application has the following advantages compared with the prior art: The initial financial public opinion data and traditional financial data are collected in real time through the multi-source heterogeneous data collection layer in the embodiments of the present application; the initial financial public opinion data is subjected to multi-language nested sentiment analysis according to the dynamic sentiment quantification engine to obtain the text sentiment analysis result, the picture sentiment analysis result, and the audio-visual sentiment analysis result; cross-modal sentiment resonance detection and sentiment fusion are performed according to the text sentiment analysis result, the picture sentiment analysis result, and the audio-visual sentiment analysis result to obtain the current market sentiment indicator; the current market sentiment indicator is an optimism value or a pessimism value; the information dissemination speed and scope of the initial financial public opinion data are calculated according to the spatio-temporal propagation network model; the current currency exchange rate information, the current market sentiment indicator, the information dissemination speed and scope, and the traditional financial data are input into the exchange rate change prediction model to obtain the exchange rate change of the target currency within a preset future time period. It can improve the efficiency and accuracy of financial public opinion data analysis, thereby improving the accuracy and reliability of exchange rate prediction.

[0106] In addition, the embodiments of the present application also combine the influence of the background music of the audio-visual on public opinion, use a panic correlation model to predict the panic correlation index, thereby predicting the influence of public opinion on people's emotions, and integrating it into the first audio-visual emotion result, making the analysis of the audio-visual emotion result more accurate.

[0107] Figure 3 The structural schematic diagram of the computer device provided by the embodiments of the present invention is shown. The specific implementation of the computer device is not limited in the specific embodiments of the present invention.

[0108] As Figure 3 shown, the computer device may include: a processor 302, a communication interface 304, a memory 306, and a communication bus 308.

[0109] Wherein: the processor 302, the communication interface 304, and the memory 306 complete mutual communication through the communication bus 308. The communication interface 304 is used to communicate with network elements of other devices such as clients or other servers. The processor 302 is configured to execute a program 310, and specifically may execute relevant steps in the above-mentioned method embodiments for big data prediction based on public opinion.

[0110] Specifically, the program 310 may include program code that includes computer-executable instructions.

[0111] The processor 302 may be a central processing unit (CPU), or a specific application integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention. One or more processors included in the computer device may be of the same type of processor, such as one or more CPUs; or may be of different types of processors, such as one or more CPUs and one or more ASICs.

[0112] The memory 306 is used to store the program 310. The memory 306 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk memory.

[0113] The program 310 can be specifically called by the processor 302 to cause the computer device to perform the following operations:

[0114] Initial financial public opinion data and traditional financial data collected in real time through the multi-source heterogeneous data collection layer; the public opinion data includes text data, picture data, and audio-visual data of multiple social media, news websites, forums, policy text data, and video platforms;

[0115] Perform multi-language nested sentiment analysis on the initial financial public opinion data according to the dynamic sentiment quantification engine to obtain text sentiment analysis results, picture sentiment analysis results, and audio-visual sentiment analysis results;

[0116] Perform cross-modal sentiment resonance detection and sentiment fusion based on the text sentiment analysis results, the picture sentiment analysis results, and the audio-visual sentiment analysis results to obtain the current market sentiment index; the current market sentiment index is an optimism value or a pessimism value;

[0117] Calculate the information propagation speed and range of the initial financial public opinion data according to the spatio-temporal propagation network model;

[0118] Input the current currency exchange rate information, the current market sentiment index, the information propagation speed and range, and the traditional financial data into the exchange rate change prediction model to obtain the exchange rate change of the target currency within a preset future time period.

[0119] In an alternative manner, the performing multi-language nested sentiment analysis on the initial financial public opinion data according to the dynamic sentiment quantification engine to obtain text sentiment analysis results, picture sentiment analysis results, and audio-visual sentiment analysis results includes:

[0120] In the BERT model and the explicit emotion recognition model, the explicit emotion corresponding to the text data is obtained; the text emotion recognition model is trained based on explicit text samples and explicit emotion labels for the LSTM model;

[0121] The TransR model is used to calculate the semantic distance between the text data in the initial financial public opinion data and the metaphor knowledge graph, and determine the metaphor emotion of the text data; the metaphor knowledge graph includes multiple nodes, and each node is a financial metaphor triple, and the elements of the financial metaphor triple are metaphor text, corresponding explicit meaning and relationship;

[0122] According to the preset dynamic dictionary and the attention weight analysis algorithm, the text data in the initial financial public opinion data is analyzed to obtain policy expectation game information; the preset dynamic dictionary is constructed in advance based on policy vocabulary;

[0123] According to the explicit emotion, metaphor emotion signal and policy expectation game emotion information, the text emotion analysis result is determined.

[0124] In an optional manner, in the BERT model and the explicit emotion recognition model, obtaining the explicit emotion corresponding to the text data further includes:

[0125] The BERT model is used to encode the text data in the initial financial public opinion data and extract the context embedding vector;

[0126] The context embedding vector is input into the explicit emotion recognition model to obtain the explicit emotion corresponding to the target currency in the text data; the text emotion recognition model is trained based on explicit text samples and explicit emotion labels for the LSTM model.

[0127] In an optional manner, in using the TransR model to calculate the semantic distance between the text data in the initial financial public opinion data and the metaphor knowledge graph and determine the metaphor emotion of the text data, the method includes:

[0128] The text data in the initial financial public opinion data is matched with the metaphor triples in the metaphor knowledge graph to obtain the matching text in the text data;

[0129] The TransR model is used to calculate the semantic distance between the matching text and the metaphor knowledge graph; among them, the TransR model is iteratively trained according to the preset text samples according to the preset loss function; the preset loss function is:

[0130]

[0131] Among them, S metaphor is the loss function; h r= M r ·h is the mapping of the metaphorical text h under the relationship r; t r = M r ·t is the mapping of the explicit meaning t under the relationship r; h is the metaphorical text in the metaphor triple, r is the relationship in the metaphor triple, and t is the explicit meaning in the metaphor triple;

[0132] When the semantic distance is less than the preset distance, the explicit meaning corresponding to the matching text is obtained;

[0133] Determine the metaphorical emotion corresponding to the text data according to the explicit meaning.

[0134] In an alternative way, the cross-modal emotion resonance detection is performed according to the text emotion analysis result, the picture emotion analysis result, and the audio-visual emotion analysis result to obtain the current market emotion index, including:

[0135] Align the text emotion analysis result, the picture emotion analysis result, and the audio-visual emotion analysis result to the same feature space;

[0136] Calculate the similarity between the text emotion analysis result and the picture emotion analysis result and the audio-visual emotion analysis result respectively. When there is an emotion conflict, determine the authenticity of the initial financial public opinion data;

[0137] When there is no emotion conflict, perform weighted fusion calculation on the text emotion analysis result, the picture emotion analysis result, and the audio-visual emotion analysis result to obtain the current market emotion index.

[0138] In an alternative way, before the multi-language nested emotion analysis of the initial financial public opinion data is performed according to the dynamic emotion quantification engine to obtain the text emotion analysis result, the picture emotion analysis result, and the audio-visual emotion analysis result, the method includes:

[0139] Extract the text content corresponding to the audio-visual data in the initial financial public opinion data;

[0140] Extract the background music rhythm corresponding to the audio-visual data in the initial financial public opinion data;

[0141] Determine the audio-visual emotion analysis result according to the text content and the background music rhythm.

[0142] In an alternative way, the determining the audio-visual emotion analysis result according to the text content and the background music rhythm includes:

[0143] Encode the text content through the BERT model and input it into the audio-visual emotion recognition model to obtain the first audio-visual emotion corresponding to the text content;

[0144] Input the background music rhythm into the panic correlation model to obtain a panic correlation index; wherein, the panic correlation model is obtained by pre - modeling the correlation between the historical short - video background music rhythm and the market panic index.

[0145] Adjust the weight of the first audio - video emotion according to the panic correlation index to obtain the audio - video emotion analysis result.

[0146] The above - mentioned technical solution provided by the embodiments of the present application has the following advantages compared with the prior art: The embodiments of the present application collect initial financial public opinion data and traditional financial data in real - time through a multi - heterogeneous data collection layer; perform multi - language nested sentiment analysis on the initial financial public opinion data according to a dynamic sentiment quantification engine to obtain text emotion analysis results, picture emotion analysis results, and audio - video emotion analysis results; perform cross - modal emotion resonance detection and emotion fusion according to the text emotion analysis results, the picture emotion analysis results, and the audio - video emotion analysis results to obtain the current market emotion index; the current market emotion index is an optimism value or a pessimism value; calculate the information dissemination speed and scope of the initial financial public opinion data according to a spatio - temporal propagation network model; input the current currency exchange rate information, the current market emotion index, the information dissemination speed and scope, and the traditional financial data into an exchange rate change prediction model to obtain the exchange rate change of the target currency within a preset future time period. It can improve the efficiency and accuracy of financial public opinion data analysis, thereby improving the accuracy and reliability of exchange rate prediction.

[0147] In addition, the embodiments of the present application also consider the impact of the background music of audio - visual content on public opinion. By using a panic correlation model, the panic correlation index is predicted, so as to predict the impact of public opinion on people's emotions and fuse it into the first audio - video emotion result, making the analysis of the audio - video emotion result more accurate.

[0148] The embodiments of the present invention provide a computer - readable storage medium, and the storage medium stores at least one executable instruction. When the executable instruction runs on a computer device, the computer device is enabled to execute the big - data prediction method based on public opinion in any of the above - mentioned method embodiments.

[0149] The executable instruction can specifically be used to enable the computer device to perform the following operations:

[0150] Collect initial financial public opinion data and traditional financial data in real - time through a multi - heterogeneous data collection layer; the public opinion data includes text data, picture data, and audio - video data from multiple social media, news websites, forums, policy text data, and video platforms.

[0151] Perform multi - language nested sentiment analysis on the initial financial public opinion data according to the dynamic sentiment quantification engine to obtain text sentiment analysis results, picture sentiment analysis results, and audio - video sentiment analysis results;

[0152] Perform cross - modal emotion resonance detection and emotion fusion based on the text sentiment analysis results, the picture sentiment analysis results, and the audio - video sentiment analysis results to obtain the current market emotion index; the current market emotion index is an optimism value or a pessimism value;

[0153] Calculate the information dissemination speed and scope of the initial financial public opinion data according to the spatio - temporal propagation network model;

[0154] Input the current currency exchange rate information, the current market emotion index, the information dissemination speed and scope, and the traditional financial data into the exchange rate change prediction model to obtain the exchange rate change of the target currency within a preset future time period.

[0155] In an alternative approach, the performing multi - language nested sentiment analysis on the initial financial public opinion data according to the dynamic sentiment quantification engine to obtain text sentiment analysis results, picture sentiment analysis results, and audio - video sentiment analysis results includes:

[0156] Obtain the explicit emotion corresponding to the text data through the BERT model and the explicit emotion recognition model; the text emotion recognition model is trained based on explicit text samples and explicit emotion labels for the LSTM model;

[0157] Use the TransR model to calculate the semantic distance between the text data in the initial financial public opinion data and the metaphor knowledge graph, and determine the metaphor emotion of the text data; the metaphor knowledge graph includes multiple nodes, and each node is a financial metaphor triple, and the elements of the financial metaphor triple are metaphor text, corresponding explicit meaning, and relationship;

[0158] Analyze the text data in the initial financial public opinion data according to the preset dynamic dictionary and attention weight analysis algorithm to obtain policy expectation game information; the preset dynamic dictionary is constructed in advance based on policy vocabulary;

[0159] Determine the text sentiment analysis result according to the explicit emotion, metaphor emotion signal, and policy expectation game emotion information.

[0160] In an alternative approach, the obtaining the explicit emotion corresponding to the text data through the BERT model and the explicit emotion recognition model further includes:

[0161] Encode the text data in the initial financial public opinion data through the BERT model to extract context embedding vectors;

[0162] Input the context embedding vector into the explicit emotion recognition model to obtain the explicit emotion corresponding to the target currency in the text data; the text emotion recognition model is trained based on explicit text samples and explicit emotion labels for the LSTM model.

[0163] In an alternative approach, to calculate the semantic distance between the text data in the initial financial public opinion data and the metaphor knowledge graph using the TransR model and determine the metaphor emotion of the text data, the method includes:

[0164] Match the text data in the initial financial public opinion data with the metaphor triples of the metaphor knowledge graph to obtain the matching text in the text data;

[0165] Use the TransR model to calculate the semantic distance between the matching text and the metaphor knowledge graph; where the TransR model is iteratively trained according to preset text samples based on a preset loss function; the preset loss function is:

[0166]

[0167] where, S metaphor is the loss function; h r = M r ·h is the mapping of the metaphorical text h under the relation r; t r = M r ·t is the mapping of the explicit meaning t under the relation r; h is the metaphorical text in the metaphor triple, r is the relation in the metaphor triple, and t is the explicit meaning in the metaphor triple;

[0168] When the semantic distance is less than the preset distance, obtain the explicit meaning corresponding to the matching text;

[0169] Determine the metaphor emotion corresponding to the text data based on the explicit meaning.

[0170] In an alternative approach, for cross-modal emotion resonance detection based on the text emotion analysis result, the picture emotion analysis result, and the audio-visual emotion analysis result to obtain the current market emotion indicator, it includes:

[0171] Align the text emotion analysis result, the picture emotion analysis result, and the audio-visual emotion analysis result to the same feature space;

[0172] Calculate the similarity between the text emotion analysis result and the picture emotion analysis result and the audio-visual emotion analysis result respectively. When there is an emotion conflict, determine the authenticity of the initial financial public opinion data;

[0173] When the emotions do not conflict, perform weighted fusion calculation on the text emotion analysis result, the picture emotion analysis result, and the audio-visual emotion analysis result to obtain the current market emotion index.

[0174] In an alternative way, before obtaining the text emotion analysis result, the picture emotion analysis result, and the audio-visual emotion analysis result by performing multi-language nested emotion analysis on the initial financial public opinion data according to the dynamic emotion quantification engine, the method includes:

[0175] Extract the text content corresponding to the audio-visual data in the initial financial public opinion data;

[0176] Extract the background music rhythm corresponding to the audio-visual data in the initial financial public opinion data;

[0177] Determine the audio-visual emotion analysis result according to the text content and the background music rhythm.

[0178] In an alternative way, the determining the audio-visual emotion analysis result according to the text content and the background music rhythm includes:

[0179] Encode the text content through a BERT model and input it into an audio-visual emotion recognition model to obtain the first audio-visual emotion corresponding to the text content;

[0180] Input the background music rhythm into a panic correlation model to obtain a panic correlation index; wherein, the panic correlation model is pre-established by performing correlation modeling on the historical short video background music rhythm and the market panic index;

[0181] Adjust the weight of the first audio-visual emotion according to the panic correlation index to obtain the audio-visual emotion analysis result.

[0182] The above technical solution provided by the embodiments of the present application has the following advantages compared with the prior art: The embodiments of the present application collect initial financial public opinion data and traditional financial data in real time through a multi-source heterogeneous data collection layer; perform multi-language nested sentiment analysis on the initial financial public opinion data according to a dynamic sentiment quantification engine to obtain text sentiment analysis results, picture sentiment analysis results, and audio-visual sentiment analysis results; perform cross-modal sentiment resonance detection and sentiment fusion according to the text sentiment analysis results, the picture sentiment analysis results, and the audio-visual sentiment analysis results to obtain the current market sentiment index; the current market sentiment index is an optimism value or a pessimism value; calculate the information dissemination speed and scope of the initial financial public opinion data according to a spatio-temporal propagation network model; input the current currency exchange rate information, the current market sentiment index, the information dissemination speed and scope, and the traditional financial data into an exchange rate change prediction model to obtain the exchange rate change of the target currency within a preset future time period. It can improve the efficiency and accuracy of financial public opinion data analysis, thereby improving the accuracy and reliability of exchange rate prediction.

[0183] In addition, the embodiments of the present application also consider the impact of the background music of the audio-visual on public opinion. By using a panic correlation model, the panic correlation index is predicted, so as to predict the impact of public opinion on people's emotions, and fuse it into the first audio-visual emotion result, making the analysis of the audio-visual emotion result more accurate.

[0184] The embodiments of the present invention provide a big data prediction device based on public opinion for executing the above-mentioned big data prediction method based on public opinion.

[0185] The embodiments of the present invention provide a computer program, which can be called by a processor to enable a computer device to execute the big data prediction method based on public opinion in any of the above method embodiments.

[0186] The embodiments of the present invention provide a computer program product. The computer program product includes a computer program stored on a computer-readable storage medium. The computer program includes program instructions. When the program instructions run on a computer, the computer is enabled to execute the big data prediction method based on public opinion in any of the above method embodiments.

[0187] The algorithms or displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. The structure required to construct such a system will be apparent from the above description. In addition, the embodiments of the present invention are not directed to any specific programming language. It should be understood that the content of the present invention described herein can be implemented using various programming languages, and the description of a specific language above is to disclose the best implementation mode of the present invention.

[0188] In the description provided herein, numerous specific details are set forth. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure an understanding of this description.

[0189] Similarly, it should be understood that in order to streamline the present invention and assist in understanding one or more of the various inventive aspects, in the foregoing description of exemplary embodiments of the present invention, various features of the embodiments of the present invention are sometimes grouped together in a single embodiment, figure, or description thereof. However, the disclosed method should not be construed as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim.

[0190] Those skilled in the art will appreciate that the modules in the devices in the embodiments can be adaptively changed and disposed in one or more devices different from the embodiments. The modules or units or components in the embodiments can be combined into one module or unit or component, and can be divided into multiple sub-modules or sub-units or sub-components. Except that at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all the features disclosed in this specification (including the accompanying claims, abstract and drawings) and all the processes or units of any method or device so disclosed. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstract and drawings) can be replaced by an alternative feature providing the same, equivalent or similar purpose.

[0191] It should be noted that the above embodiments illustrate rather than limit the present invention, and those skilled in the art can design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in the claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The present invention can be implemented by means of hardware including several different elements and by means of a suitably programmed computer. In the unit claims listing several devices, several of these devices can be embodied by the same item of hardware. The use of the words first, second, and third, etc. does not denote any order. These words can be interpreted as names. The steps in the above embodiments, unless otherwise specifically stated, should not be construed as limiting the order of execution.

Claims

1. A big data prediction method based on public opinion, characterized in that: The method comprises: Initial financial public opinion data and traditional financial data collected in real time through a multi-heterogeneous data collection layer; the public opinion data includes text data, image data, and audio and video data from multiple social media, news websites, forums, policy text data, and video platforms; Perform multi-language nested sentiment analysis on the initial financial public opinion data using the dynamic sentiment quantification engine to obtain text sentiment analysis results, image sentiment analysis results, and audio and video sentiment analysis results; Perform cross-modal emotion resonance detection and emotion fusion according to the text emotion analysis result, the picture emotion analysis result, and the audio and video emotion analysis result to obtain a current market emotion index; the current market emotion index is an optimism value or a pessimism value; Calculate the information propagation speed and scope of the initial financial public opinion data based on the spatiotemporal propagation network model; The current currency exchange rate information, the current market sentiment index, the information transmission speed and range, and the traditional financial data are input into the exchange rate change prediction model to obtain the exchange rate change of the target currency within a preset time period in the future.

2. The method according to claim 1, characterized in that The multi-language nested sentiment analysis is performed on the initial financial public opinion data according to the dynamic sentiment quantification engine to obtain text sentiment analysis results, image sentiment analysis results and audio and video sentiment analysis results, including: Obtaining explicit emotions corresponding to text data through the BERT model and the explicit emotion recognition model; the text emotion recognition model is obtained by training the LSTM model according to the explicit text samples and the explicit emotion labels; The TransR model is used to calculate the semantic distance between the text data in the initial financial public opinion data and the metaphor knowledge graph to determine the metaphorical sentiment of the text data; the metaphor knowledge graph includes a plurality of nodes, each node is a financial metaphor triple, and the elements of the financial metaphor triple are metaphor text, corresponding explicit meaning and relationship; According to a preset dynamic dictionary and an attention weight analysis algorithm, the text data in the initial financial public opinion data is analyzed to obtain policy expectation game information; the preset dynamic dictionary is constructed in advance based on policy vocabulary; The text sentiment analysis result is determined based on the explicit emotions, metaphorical emotion signals and policy expectation game emotion information.

3. The method according to claim 2, characterized in that The method of obtaining the explicit emotion corresponding to the text data through the BERT model and the explicit emotion recognition model further includes: Encode the text data in the initial financial public opinion data through the BERT model to extract the context embedding vector; The context embedding vector is input into an explicit emotion recognition model to obtain the explicit emotion corresponding to the target currency in the text data; the text emotion recognition model is obtained by training an LSTM model according to explicit text samples and explicit emotion labels.

4. The method according to claim 2, characterized in that: The method of using the TransR model to calculate the semantic distance between the text data in the initial financial public opinion data and the metaphor knowledge graph to determine the metaphorical emotion of the text data includes: Matching the text data in the initial financial public opinion data with the metaphor triples of the metaphor knowledge graph to obtain matching text in the text data; The semantic distance between the matching text and the metaphor knowledge graph is calculated using the TransR model; wherein the TransR model is iteratively trained according to a preset text sample and a preset loss function; the preset loss function is: Among them, S metaphor is the loss function; h r =M r h is the mapping of metaphor text h under relation r; t r =M r t is the mapping of explicit meaning t under relation r; h is the metaphor text in the metaphor triple, r is the relation in the metaphor triple, and t is the explicit meaning in the metaphor triple; When the semantic distance is less than a preset distance, obtaining the explicit meaning corresponding to the matching text; The metaphorical emotion corresponding to the text data is determined according to the explicit meaning.

5. The method according to claim 1, characterized in that The cross-modal emotion resonance detection is performed according to the text emotion analysis result, the picture emotion analysis result and the audio and video emotion analysis result to obtain the current market emotion index, including: Aligning the text emotion analysis result, the picture emotion analysis result, and the audio and video emotion analysis result to the same feature space; Calculate the similarity between the text sentiment analysis results, the picture sentiment analysis results, and the audio and video sentiment analysis results respectively, and when there is a conflict of sentiments, determine the authenticity of the initial financial public opinion data; When there is no conflict in emotions, a weighted fusion calculation is performed on the text emotion analysis result, the picture emotion analysis result and the audio and video emotion analysis result to obtain a current market emotion index.

6. The method according to any one of claims 1 to 5, characterized in that: The multi-language nested sentiment analysis is performed on the initial financial public opinion data according to the dynamic sentiment quantification engine to obtain text sentiment analysis results, image sentiment analysis results and audio and video sentiment analysis results, including: Extracting text content corresponding to the audio and video data in the initial financial public opinion data; Extracting the background music rhythm corresponding to the audio and video data in the initial financial public opinion data; The audio and video emotion analysis result is determined according to the text content and the background music rhythm.

7. The method according to claim 6, characterized in that Determining the audio and video emotion analysis result according to the text content and the background music rhythm includes: The text content is encoded by a BERT model and input into an audio and video emotion recognition model to obtain a first audio and video emotion corresponding to the text content; Inputting the background music rhythm into a panic association model to obtain a panic association index; wherein the panic association model is obtained by pre-modeling the correlation between the background music rhythm of historical short videos and the market panic index; The first audio and video emotion is weighted according to the panic association index to obtain the audio and video emotion analysis result.

8. A big data prediction platform based on public opinion, characterized in that: The platform includes: A collection module, which is used to collect initial financial public opinion data and traditional financial data in real time through a multi-heterogeneous data collection layer; the public opinion data includes text data, image data, and audio and video data from multiple social media, news websites, forums, policy text data, and video platforms; The sentiment analysis module is used to perform multi-language nested sentiment analysis on the initial financial public opinion data based on the dynamic sentiment quantification engine, and obtain text sentiment analysis results, image sentiment analysis results, and audio and video sentiment analysis results; An index calculation module, used to perform cross-modal emotion resonance detection and emotion fusion according to the text emotion analysis result, the picture emotion analysis result and the audio and video emotion analysis result, to obtain a current market emotion index; the current market emotion index is an optimism value or a pessimism value; The propagation calculation module is used to calculate the information propagation speed and scope of the initial financial public opinion data based on the spatiotemporal propagation network model; The prediction module is used to input the exchange rate change prediction model into the exchange rate change prediction model according to the current currency exchange rate information, the current market sentiment index, the information propagation speed and scope, and the traditional financial data, so as to obtain the exchange rate change of the target currency within a preset time period in the future.

9. A computer device, characterized in that: include: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform the operation of the public opinion-based big data prediction method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The storage medium stores at least one executable instruction, and when the executable instruction is executed on a computer device, the computer device executes the operation of the public opinion-based big data prediction method as described in any one of claims 1 to 7.

Images