Data prediction method and device, computer equipment and storage medium

Abstract

The invention belongs to the technical field of artificial intelligence, and relates to a data prediction method, which comprises the steps of collecting initial multi-modal data corresponding to a target industry from a specified platform; performing noise filtering on the initial multi-modal data based on a multi-level verification strategy to obtain multi-modal data; performing emotional feature extraction on the multi-modal data to obtain emotional feature data; performing emotion analysis on the emotion feature data based on a hybrid neural network to obtain an emotion analysis result; predicting the emotion analysis result based on a prediction model to obtain a market prediction result; and displaying the emotion analysis result and the market prediction result. The invention further provides a data prediction device, computer equipment and a storage medium. In addition, the invention also relates to a block chain technology, and the market prediction result can be stored in a block chain. The method can be applied to market condition analysis scenes in the field of financial science and technology, the accuracy of market condition prediction can be effectively improved, and the data accuracy of the generated market condition prediction result is ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and can be applied to the field of financial technology, in particular to a data prediction method and device, a computer device and a storage medium. BACKGROUND

[0002] In the market sentiment analysis technology in the traditional financial field, market sentiment analysis mainly relies on structured text data such as news headlines, financial reports, etc., by extracting information from these data and performing sentiment analysis to assist market trend judgment. However, this analysis method has obvious limitations, which lacks effective use of unstructured data (especially social media data), and does not filter false information in social media. Because social media data is widely sourced and has a low publishing threshold, it is filled with a large amount of noise data, and the existence of false information seriously interferes with the accuracy of market sentiment analysis.

[0003] This inaccurate market sentiment analysis further leads to a large deviation in market trend prediction results. For example, in the stock market trend prediction scenario, the traditional method may only determine that the stock price of a company will rise according to the positive news mentioned in the news report, but ignores the potential impact of a large amount of negative false information about the company on investors' emotions on social media. Finally, due to the lack of comprehensive and accurate analysis of market sentiment, the prediction of the stock price trend of the company fails, which makes investors face the risk of decision-making and reduces the resource allocation efficiency of the financial market.

[0004] Therefore, there is an urgent need to provide a new technology that can accurately analyze market sentiment to improve the accuracy of market trend prediction and provide strong support for the stable operation of the financial market and the rational decision-making of investors. SUMMARY

[0005] The purpose of the embodiments of the present application is to provide a data prediction method, device, computer device and storage medium to solve the technical problem of low accuracy of market trend prediction in existing market sentiment analysis technology.

[0006] In a first aspect, a data prediction method is provided, comprising:

[0007] Collecting initial multi-modal data corresponding to a target industry from a specified platform; wherein the initial multi-modal data includes text data, image data and audio data;

[0008] Filtering noise from the initial multi-modal data based on a preset multi-level verification strategy to obtain corresponding multi-modal data;

[0009] Performing sentiment feature extraction on the multi-modal data to obtain corresponding sentiment feature data;

[0010] performing emotion analysis on the emotion feature data based on a preset hybrid neural network to obtain a corresponding emotion analysis result;

[0011] performing prediction processing on the emotion analysis result based on a preset prediction model to obtain a corresponding market trend prediction result;

[0012] performing display processing on the emotion analysis result and the market trend prediction result.

[0013] In a second aspect, a data prediction apparatus is provided, comprising:

[0014] a collection module configured to collect initial multi-modal data corresponding to a target industry from a designated platform, wherein the initial multi-modal data comprises text data, image data, and audio data;

[0015] a filtering module configured to perform noise filtering on the initial multi-modal data based on a preset multi-level verification strategy to obtain corresponding multi-modal data;

[0016] an extraction module configured to perform emotion feature extraction on the multi-modal data to obtain corresponding emotion feature data;

[0017] an analysis module configured to perform emotion analysis on the emotion feature data based on a preset hybrid neural network to obtain a corresponding emotion analysis result;

[0018] a prediction module configured to perform prediction processing on the emotion analysis result based on a preset prediction model to obtain a corresponding market trend prediction result;

[0019] a display module configured to perform display processing on the emotion analysis result and the market trend prediction result.

[0020] In a third aspect, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the above data prediction method when executing the computer program.

[0021] In a fourth aspect, a computer readable storage medium is provided, and the computer readable storage medium stores a computer program, and the computer program implements the steps of the above data prediction method when executed by a processor.

[0022] In the above-mentioned data prediction method, apparatus, computer equipment, and storage medium, the following steps are taken: First, initial multimodal data corresponding to the target industry is collected from a designated platform; wherein, the initial multimodal data includes text data, image data, and audio data; then, noise filtering is performed on the initial multimodal data based on a preset multi-level verification strategy to obtain the corresponding multimodal data; next, sentiment features are extracted from the multimodal data to obtain the corresponding sentiment feature data; subsequently, sentiment analysis is performed on the sentiment feature data based on a preset hybrid neural network to obtain the corresponding sentiment analysis results; further, the sentiment analysis results are processed for prediction based on a preset prediction model to obtain the corresponding market prediction results; finally, the sentiment analysis results and the market prediction results are displayed. Based on the above automated processing flow, this application collects initial multimodal data, including text, image, and audio data, corresponding to the target industry from a designated platform. It then intelligently filters noise from the initial multimodal data using a multi-level verification strategy to obtain multimodal data. Next, it extracts sentiment features from the multimodal data to obtain sentiment feature data. Following this, it performs sentiment analysis on the sentiment feature data using a hybrid neural network to obtain sentiment analysis results. Finally, it uses a predictive model to predict the sentiment analysis results to obtain market forecast results. Finally, it displays the sentiment analysis results and market forecast results. This application effectively filters false information from multimodal data using a multi-level verification strategy. Furthermore, by combining a hybrid neural network with a predictive model to predict the sentiment feature data related to the filtered multimodal data, it effectively improves the accuracy of market forecasts and ensures the accuracy of the generated market forecast results. Attached Figure Description

[0023] To more clearly illustrate the solutions in this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is an exemplary system architecture diagram to which this application can be applied;

[0025] Figure 2 This is a flowchart of an embodiment of the data prediction method according to this application;

[0026] Figure 3 This is a schematic diagram of the structure of one embodiment of the data prediction device according to this application;

[0027] Figure 4is a structural schematic diagram of one embodiment of the computer device according to the present application. DETAILED DESCRIPTION

[0028] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terminology used in the description herein is for describing particular embodiments only and is not intended to be limiting of the application; the description herein and the claims of the application and the above description of the drawings herein are not to be construed as limiting upon the scope of the application; the description herein and the claims of the application and the above description of the drawings herein use the term "comprising" and "having" and any variations thereof to mean "including but not limited to".

[0029] Reference herein to "embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase that the phrase in the specification appear not necessarily all refer to the same embodiment, or are mutually exclusive or alternative embodiments.

[0030] In order to make the technical personnel in the art better understand the scheme of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings.

[0031] As shown in Figure 1 The system architecture 100 can include a terminal device 101, a network 102 and a server 103, and the terminal device 101 can be a notebook computer 1011, a tablet computer 1012 or a mobile phone 1013. The network 102 is a medium for providing a communication link between the terminal device 101 and the server 103. The network 102 can include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0032] A user can use the terminal device 101 to interact with the server 103 through the network 102 to receive or send messages, etc. Various communication client applications can be installed on the terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0033] The terminal device 101 can be various electronic devices with a display screen and supporting web browsing, in addition to the notebook computer 1011, the tablet computer 1012 or the mobile phone 1013, the terminal device 101 can also be an electronic book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 player (Moving Picture Experts Group Audio Layer IV), a laptop computer and a desktop computer, etc.

[0034] The server 103 can be a server providing various services, for example, a background server providing support for a page displayed on the terminal device 101.

[0035] It should be noted that the data prediction method provided by the embodiments of the present application is generally executed by a server / terminal device, and accordingly, the data prediction apparatus is generally arranged in a server / terminal device.

[0036] It should be understood that Figure 1 The number of terminal devices, networks and servers in

[0037] With reference to Figure 2 , a flowchart of one embodiment of the data prediction method according to the present application is shown. The order of the steps in the flowchart can be changed according to different needs, and some steps can be omitted. The data prediction method provided by the embodiments of the present application can be applied to any scenario requiring market trend prediction, and then the data prediction method can be applied to the products in these scenarios, for example, market trend prediction in the financial field. The data prediction method comprises the following steps:

[0038] Step S201, collecting initial multi-modal data corresponding to a target industry from a specified platform; wherein the initial multi-modal data comprises text data, image data and audio data.

[0039] In the present embodiment, the electronic device (for example Figure 1The server / terminal device shown) can obtain initial multi-modal data corresponding to the target industry through wired or wireless connection. It should be noted that the wireless connection can include but is not limited to 3G / 4G / 5G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other now known or future developed wireless connection. The subject of the present application is a data prediction system, also known as a market prediction system, which can be referred to as a system. The target industry can include the encryption asset industry or the pharmaceutical industry, and can also include other related industries. The specified platform can include at least social media platforms, related financial websites and video platforms, and financial report conference call audio sources. The data collection process of the initial multi-modal data includes organizing multiple servers into a cluster, and running a crawler program on each server. Through a distributed architecture, data can be collected concurrently from multiple data sources, improving collection efficiency and speed. Then, due to the limitations of API call frequency on multiple platforms, by using dynamic IP camouflage technology, a different IP address is randomly assigned each time a request is initiated, simulating the access behavior of different users, breaking through the call frequency limit of the platform, and ensuring that data can be continuously and stably collected from these platforms. Further, text data is collected from social media platforms, including user posts, comments, etc.; image data containing market information, such as meme graphs and K-line graphs, is collected from related financial websites and video platforms; and audio data is collected from financial report conference call audio sources.

[0040] Step S202, based on a preset multi-level verification strategy, noise filtering is performed on the initial multi-modal data to obtain corresponding multi-modal data.

[0041] In this embodiment, the specific implementation process of the above-mentioned based on a preset multi-level verification strategy, noise filtering is performed on the initial multi-modal data to obtain corresponding multi-modal data will be further described in detail in the subsequent specific embodiments, which will not be described here.

[0042] After obtaining the multi-modal data, the timestamp asynchronous problem of cross-platform data can be processed by calling a specially constructed spatio-temporal alignment module. The spatio-temporal alignment module first collects the timestamp information of multi-modal data from different platforms, and then adjusts the timestamps of data from different sources to a unified time reference according to the generation time and transmission delay of the data, etc., using time interpolation, synchronization calibration and other algorithms, to ensure that the time error is controlled within 50 ms. At the same time, for data involving geographic location, the data is aligned in the spatial dimension according to the source and collection time of the data, to ensure that data from different sources can accurately correspond in time and space dimensions, to obtain the final required multi-modal data. The spatio-temporal alignment module solves the inconsistency of time and space of cross-platform data, ensuring that the subsequent steps can accurately fuse and analyze these multi-modal data, so that the entire analysis process is based on a unified time and space reference.

[0043] In step S203, the multi-modal data is subjected to emotion feature extraction to obtain corresponding emotion feature data.

[0044] In the present embodiment, the process of the above-mentioned emotion feature extraction includes: for the text layer, the collected text data is input into the DeBERTa-V3 model. The model is pre-trained with a large amount of text data and can perform fine-grained sentiment analysis on the text. By setting a 9-level emotion intensity classification standard, the model can output the emotion intensity level expressed by each sentence or paragraph in the text, thereby accurately capturing the emotional information in the text. For the visual layer, the image data is input into the CLIP model. The CLIP model has strong image-text matching capability and can understand the semantic information in the image. For meme images, it can identify the humor, satire and other emotions conveyed by the combination of text and images in the image; for K-line charts, it can analyze the market sentiment signals implied by the trend, pattern and other characteristics of the chart, such as the upward trend may indicate optimistic sentiment, etc. For the audio layer: the audio data of the earnings conference call is input into the Whisper-LLM architecture. The Whisper model is good at speech recognition and can convert audio to text. The LLM (Large Language Model) part can perform semantic understanding on the converted text, and at the same time, combined with the features such as tone and speed of the audio, extract the emotional fluctuation features of the speaker in the earnings conference call, such as excitement, nervousness, etc. Then, the emotion features extracted from the text layer, visual layer and audio layer are integrated to obtain the corresponding initial emotion feature data.

[0045] In step S204, the emotion feature data is subjected to emotion analysis based on a pre-set hybrid neural network to obtain a corresponding emotion analysis result.

[0046] In the embodiment, the specific implementation process of performing emotion analysis on the emotion feature data based on the preset hybrid neural network to obtain a corresponding emotion analysis result will be further described in detail in subsequent specific embodiments, and will not be described in detail here.

[0047] In step S205, the emotion analysis result is processed based on a preset prediction model to obtain a corresponding market trend prediction result.

[0048] In the embodiment, the emotion analysis result can be input into a pre-constructed prediction model, the prediction model can analyze the emotion analysis result according to the mapping relationship between the learned market emotion features and market trends (such as VIX index, stock amplitude, etc.), and output a prediction result of the market trend, that is, the market trend prediction result. The specific construction process of the prediction model will be further described in detail in subsequent specific embodiments, and will not be described in detail here.

[0049] In step S206, the emotion analysis result and the market trend prediction result are displayed.

[0050] In the embodiment, the specific implementation process of displaying the emotion analysis result and the market trend prediction result will be further described in detail in subsequent specific embodiments, and will not be described in detail here.

[0051] The application first collects initial multi-modal data corresponding to the target industry from a designated platform; wherein the initial multi-modal data includes text data, image data and audio data; then performs noise filtering on the initial multi-modal data based on a preset multi-level verification strategy to obtain corresponding multi-modal data; then extracts emotional features from the multi-modal data to obtain corresponding emotional feature data; subsequently performs emotional analysis on the emotional feature data based on a preset hybrid neural network to obtain corresponding emotional analysis results; further performs prediction processing on the emotional analysis results based on a preset prediction model to obtain corresponding market trend prediction results; and finally performs display processing on the emotional analysis results and the market trend prediction results. Based on the above automatic processing procedure, the application collects initial multi-modal data including text data, image data and audio data corresponding to the target industry from a designated platform, and intelligently performs noise filtering on the initial multi-modal data based on the use of a multi-level verification strategy to obtain multi-modal data, then extracts emotional features from the multi-modal data to obtain emotional feature data, subsequently performs emotional analysis on the emotional feature data based on the use of a hybrid neural network to obtain emotional analysis results, further performs prediction processing on the emotional analysis results based on the use of a prediction model to obtain market trend prediction results, and finally performs display processing on the emotional analysis results and the market trend prediction results. The application effectively filters false information from multi-modal data based on a multi-level verification strategy, and further performs prediction processing on emotional feature data related to the filtered multi-modal data based on the combined use of a hybrid neural network and a prediction model, which can effectively improve the accuracy of market trend prediction and ensure the data accuracy of the generated market trend prediction results.

[0052] In some optional implementations, the multi-level verification strategy includes an account-level verification strategy, a content-level verification strategy and a propagation-level verification strategy; step S202 includes the following steps:

[0053] Perform noise filtering on the initial multi-modal data based on the account-level verification strategy to obtain corresponding first processing data.

[0054] In this embodiment, the strategy content of the above-mentioned account-level verification strategy includes collecting data such as registration time of a user account, posting mode (such as posting frequency, posting time distribution, etc.), device fingerprint (unique identification information of a device), etc. By analyzing these data, a botnet identification model is established, and the initial multi-modal data is filtered based on the botnet identification model to obtain corresponding first processing data. For example, if an account has a very short registration time, an abnormal posting mode (such as a large number of similar content published in a short period of time), and a device fingerprint identical to that of other multiple accounts, the account is likely to be a bot account, and the data generated by the account is filtered out to improve the authenticity and reliability of the data.

[0055] Noise filtering is performed on the first processing data based on the content level verification strategy to obtain corresponding second processing data.

[0056] In the embodiment, the policy content of the content level verification strategy includes: using a Grover-Transformer model to detect AI-generated text existing in the initial multi-modal data, and deleting the AI-generated text to obtain corresponding second processing data. The Grover-Transformer is trained by a large amount of real text and AI-generated text, and can identify the differences in language features and logical structure of the text. The collected text data is input into the model, and the model outputs a probability value of the text being AI-generated. When the probability value exceeds a certain threshold, it is determined that the text is AI-generated content, and these data are excluded to avoid interference with market sentiment analysis.

[0057] Noise filtering is performed on the second processing data based on the propagation level verification strategy to obtain corresponding third processing data.

[0058] In the embodiment, the policy content of the propagation level verification strategy includes: analyzing the topology structure of the information diffusion network, including the degree distribution, clustering coefficient, and shortest path of the nodes. By monitoring the changes of these indicators, abnormal propagation clusters are identified. For example, if an information rapidly spreads from a small node cluster to a large number of nodes in a short time, and the connection relationship between these nodes is abnormally close, the information propagation cluster may be abnormal. By discovering and excluding these abnormally propagated data, the data environment is further purified.

[0059] The third processing data is used as the multi-modal data.

[0060] In the embodiment, the system can also construct a dynamic credibility scoring model that considers information at the account level, content level, and propagation level. A corresponding weight is set for the indicators of each level, and a credibility score in the interval of 0-1 is calculated according to the performance of the data at different levels. For example, if the data comes from an account with a long registration time, a normal posting pattern, and a unique device fingerprint, the content is determined to be non-AI-generated, and the propagation process is normal, then the credibility score of the data will be higher. The credibility score is used as a weighting coefficient for the sentiment indicator to adjust the sentiment indicator, so that the final sentiment analysis result is more accurate and reliable.

[0061] The application filters noise from the initial multi-modal data based on the account-level verification strategy to obtain corresponding first processed data; then filters noise from the first processed data based on the content-level verification strategy to obtain corresponding second processed data; then filters noise from the second processed data based on the propagation-level verification strategy to obtain corresponding third processed data; and subsequently uses the third processed data as the multi-modal data. Based on the above processing procedure, the application identifies and excludes noise data in the initial multi-modal data from three levels of account, content and propagation based on the use of the multi-level verification strategy, thereby ensuring the data accuracy of the obtained multi-modal data and improving the quality of the multi-modal data used in subsequent steps of analysis.

[0062] In some optional implementations of the embodiment, the hybrid neural network includes a graph convolution network and a time convolution network; and step S204 includes the following steps:

[0063] Performing emotion analysis on the emotion feature data based on the graph convolution network to obtain corresponding first emotion information.

[0064] In the embodiment, the emotion analysis based on the graph convolution network (GCN) includes: constructing a user social relationship graph, taking users on a social media as nodes and relationships such as attention and interaction between users as edges. The emotion feature data is taken as the attribute of the node. The GCN captures the user social influence propagation path by performing convolution operation on the graph structure. Specifically, if a user with high influence publishes a post with positive emotion, the GCN can analyze how the emotion is propagated to other users through social relationships, thereby mining market emotion information from the social level and taking it as the corresponding first emotion information.

[0065] Performing emotion analysis on the emotion feature data based on the time convolution network to obtain corresponding second emotion information.

[0066] In the embodiment, the emotion analysis based on the time convolution network (TCN) includes: inputting emotion feature data arranged in time sequence (such as text emotion intensity and image emotion signal at different time points) into the TCN. The TCN extracts market emotion cycle features by performing sliding convolution operation on the time axis through a convolution kernel, and takes them as the corresponding second emotion information. For example, by analyzing market emotion data within a period of time, the TCN can find the fluctuation rule of market emotion every week or every month, and grasp the trend of change of market emotion over time.

[0067] Performing fusion processing on the first emotion information and the second emotion information to obtain corresponding fusion information.

[0068] In this embodiment, the first emotional information and the second emotional information can be fused by splicing or weighted fusion to obtain the corresponding fused information as the corresponding emotion analysis result.

[0069] The fused information is used as the result of the sentiment analysis.

[0070] In this embodiment, a gating attention module can also be designed into the hybrid neural network. This gating attention module can dynamically calculate the weights of different multimodal data (text, images, audio) according to the needs of the current analysis scenario. For example, when analyzing the impact of a specific event on market sentiment, if the event is mainly disseminated through text information, the gating attention mechanism will assign higher weights to text modal data, so that the hybrid neural network can more effectively utilize multimodal data for sentiment modeling.

[0071] This application performs sentiment analysis on the sentiment feature data based on the graph convolutional network to obtain corresponding first sentiment information; and performs sentiment analysis on the sentiment feature data based on the temporal convolutional network to obtain corresponding second sentiment information; then fuses the first sentiment information and the second sentiment information to obtain corresponding fused information; subsequently, the fused information is used as the sentiment analysis result. This application mines market sentiment information from multiple perspectives by using a hybrid neural network. The graph convolutional network considers the influence of user relationships on sentiment propagation in social networks, the temporal convolutional network grasps the changing patterns of market sentiment over time, and fuses the first sentiment information output by the graph convolutional network and the second sentiment information output by the temporal convolutional network to generate a sentiment analysis result. This allows the hybrid neural network to analyze market sentiment more comprehensively and accurately, providing accurate sentiment analysis results for subsequent market forecasting.

[0072] In some alternative implementations, step S206 includes the following steps:

[0073] Invoke the preset counterfactual reasoning module.

[0074] In this embodiment, the counterfactual reasoning module described above has the functions of simulating parallel market states and quantifying the marginal contribution of sentiment factors. Specifically, simulating parallel market states includes: constructing a parallel market state model for a specific event (such as a Federal Reserve interest rate hike). This parallel market state model assumes that the event has not occurred, keeps other market conditions unchanged, and simulates the possible state of market sentiment and market conditions under such circumstances based on historical data and market patterns. For example, by analyzing market data from similar past economic environments where interest rate hikes did not occur, it predicts the changes in indicators such as the VIX index (Chicago Board Options Exchange Volatility Index) and individual stock prices under current market conditions if interest rate hikes do not occur.

[0075] The quantification of the marginal contribution degree of the emotional factor includes: using the Shapley value decomposition method to analyze the contribution degree of different emotional factors (such as text emotional intensity, image emotional signal, etc.) to the asset price change. The Shapley value decomposition considers the influence of each emotional factor on the asset price change in all possible combinations, and through calculation, the role of each emotional factor in the asset price change can be determined, thereby providing a causal deduction basis for the prediction result.

[0076] Based on the deduction processing of the market trend prediction result by the counterfactual reasoning module, corresponding causal deduction information is obtained.

[0077] In this embodiment, based on the use of the counterfactual reasoning module, by simulating the parallel market state when a specific event does not occur, quantifying the marginal contribution degree of the emotional factor to the asset price, and integrating information to obtain corresponding causal deduction information, a causal deduction basis is provided for the market trend prediction result, so that the market trend prediction not only predicts the market trend simply, but also explains the reasons for the market emotion and the change of the market trend, thereby providing more valuable reference for decision-making.

[0078] A preset visualization interface is called.

[0079] In this embodiment, the above-mentioned visualization interface is a pre-constructed three-dimensional visualization decision interface, which is developed with an immersive analysis dashboard, and can perform result display functions from the time dimension, the space dimension and the causal dimension.

[0080] Based on a preset multi-dimensional display strategy, the emotional analysis result, the market trend prediction result and the causal deduction information are displayed in the visualization interface.

[0081] In this embodiment, the above-mentioned multi-dimensional display strategy refers to a strategy for multi-dimensional result display from the time dimension, the space dimension and the causal dimension. Specifically, the result display from the time dimension includes: designing a time axis control, and the user can view the market emotion heat map and the prediction trajectory of any historical time point by sliding the time axis or inputting a specific time point. The emotional heat map represents the intensity and type of market emotion by different colors, such as red for positive emotion and blue for negative emotion, and the color depth represents the size of the emotional intensity. The prediction trajectory displays the prediction result of the market trend at different time points in the past, which facilitates the user to analyze the change trend of the market emotion over time and the historical prediction situation.

[0082] The result display in the spatial dimension includes: using geographic information system (GIS) technology, superimposing the institutional investor emotion density cloud map on the geographic map. According to the geographic location and emotion data of the institutional investors, the emotion density of the institutional investors in different regions is calculated, and the cloud map with different colors and transparencies is used to represent it. The deeper the color and the lower the transparency, the more concentrated and stronger the emotion of the institutional investors in the region, which shows the emotion distribution of the institutional investors in different regions from the spatial perspective.

[0083] The result display in the causal dimension includes: using graphical method to display the Granger causality network of the emotion indicators and the market indicators. The emotion indicators and the market indicators are taken as nodes, if there is a Granger causality between two indicators, a directed edge is used to connect the two nodes, and the direction of the edge represents the direction of the causality. In this way, the causal relationship between the emotion indicators and the market indicators is intuitively presented, helping the user to understand how the market emotion changes affect the market trend.

[0084] Among them, the emotion analysis result, the market trend prediction result and the causal deduction information are correspondingly displayed in the above-mentioned visualization interface by the way of multi-dimensional result display based on the time dimension, the spatial dimension and the causal dimension contained in the above-mentioned multi-dimensional display strategy.

[0085] The present application calls a preset counterfactual reasoning module, then deduces the market trend prediction result based on the counterfactual reasoning module to obtain corresponding causal deduction information, then calls a preset visualization interface, and subsequently displays the emotion analysis result, the market trend prediction result and the causal deduction information in the visualization interface based on the preset multi-dimensional display strategy. Based on the above processing procedure, the present application deduces the causal deduction information from the market trend prediction result by using the counterfactual reasoning module, and then displays the emotion analysis result, the market trend prediction result and the causal deduction information by using the combination of the multi-dimensional display strategy and the visualization interface, which automatically and intelligently realizes the intuitive visualization of the market emotion and the market trend information obtained through complex data analysis, facilitates the user to analyze and understand from different angles, improves the intelligence of data display, and improves the user experience.

[0086] In some optional implementations, before step S205, the above-mentioned electronic device can further perform the following steps:

[0087] Obtaining the pre-constructed multi-modal sample data corresponding to the target industry.

[0088] In this embodiment, the target industry can include the encrypted asset industry in the financial field or the pharmaceutical industry, and can also include other related industries. For the encrypted asset industry in the financial field, the construction process of the multi-modal sample data includes: performing feature extraction on data such as text (such as social media discussions, news reports), images (such as market trend charts), audio (such as related audio and video descriptions) and the like related to encrypted assets. For text data, use natural language processing technology to extract emotional features; for image data and audio data, use computer vision technology to extract emotional features. Further, use a hybrid neural network to perform emotional analysis on the extracted emotional features, and use the obtained analysis results as the corresponding multi-modal sample data.

[0089] For the pharmaceutical industry, the construction process of the multi-modal sample data includes: performing emotional feature extraction on text data, image data and audio data related to the pharmaceutical industry such as social media, news reports, and using a hybrid neural network to perform emotional analysis on the extracted emotional features, and using the obtained analysis results as the corresponding multi-modal sample data. Among them, a pre-trained language model (such as BERT) can be used to encode the text to obtain the emotional vector representation of the text. For example, by analyzing a news report about the success of a drug in a phase III clinical trial, positive emotional features are extracted.

[0090] Obtain industry knowledge corresponding to the target industry, and construct corresponding mapping rules based on the industry knowledge.

[0091] In this embodiment, for the encrypted asset industry in the financial field, the construction process of the mapping rule includes: first obtaining clinical trial terms from related medical websites, cleaning, classifying and storing these terms through natural language processing technology, and constructing a special medical clinical trial term library. For example, the terms are classified and arranged according to disease types (such as cancer, cardiovascular disease, etc.), clinical trial stages (phase I, phase II, phase III, etc.). Further, collect relevant data of drug approval stages and stock price fluctuations in history, including the time when the drug enters different clinical trial stages, the approval result and the stock price data of the corresponding pharmaceutical company in the corresponding time period. Use statistical analysis methods to analyze the fluctuation rules of the stock price in different approval stages, and establish the mapping rules of the drug approval stage and the stock price fluctuation. For example, calculate the average amplitude and probability of the stock price of the corresponding pharmaceutical company rising when the drug enters the phase III clinical trial and the test result is positive.

[0092] For the pharmaceutical industry, the construction process of the above mapping rule includes: combining CoinGecko chain data, real-time monitoring of various indicators of the encrypted asset market, such as transaction volume, gas fee, NFT transaction, etc. Through API interface or other data collection tools, the data of these indicators is collected regularly. Then use data mining algorithms (such as association rule mining) to capture co-occurrence patterns such as "Gas fee surge NFT sell-off" as the corresponding mapping rule. For example, when the gas fee suddenly rises sharply, observe the transaction volume and price changes of the NFT market, if a large number of NFTs are found to be sold off, this pattern is recorded and its frequency and impact are analyzed.

[0093] A preset initial mixed model is called.

[0094] In this embodiment, the selection of the above initial mixed model is not specifically limited, and can be determined according to actual business needs, and preferably an LSTM-CNN mixed model can be used. The advantages of the LSTM-CNN mixed model include: LSTM (Long Short Term Memory Network) is good at processing sequence data and can capture long-term dependencies in data. For market trends with time series characteristics, LSTM can learn the rules of changes over time. CNN (Convolutional Neural Network) is good at extracting local features from data and performs well in processing image, text and other data. Combining the two into an LSTM-CNN mixed model can take advantage of their strengths, better handle multi-source heterogeneous market data, and extract more comprehensive and accurate features for market trend prediction.

[0095] Based on the multi-modal sample data and the mapping rule, the initial mixed model is trained to obtain a trained specified mixed model.

[0096] In this embodiment, the mapping rule and the multi-modal sample data are fused to obtain the corresponding fusion features, and then the fusion features are used as input. The historical data (i.e. fusion features) are used as a training set to train the initial mixed model, so that it learns the mapping relationship between market sentiment features and market trends (such as VIX index, stock amplitude, etc.), and the trained specified mixed model is used as the final market trend prediction model.

[0097] The specified mixed model is used as the market trend prediction model.

[0098] In the actual application, the real-time acquired data related to the medical industry (such as the progress of new clinical trials, social media sentiment, etc.) is input into the trained market trend prediction model, and the market trend prediction model outputs the prediction result of the market trend according to the learned mapping relationship, such as predicting the amplitude range of the stock price of a medical enterprise in a future period of time. Or, the real-time acquired data related to the encryption asset (such as the change of new transaction indicators, social media sentiment, etc.) is input into the trained market trend prediction model, and the market trend prediction model outputs the prediction result of the market trend according to the learned mapping relationship, such as predicting the price trend of a certain encryption asset in a future period of time.

[0099] The application obtains the pre-constructed multi-modal sample data corresponding to the target industry; then obtains the industry knowledge corresponding to the target industry, and constructs the corresponding mapping rule based on the industry knowledge; then calls the preset initial mixed model; and trains the initial mixed model based on the multi-modal sample data and the mapping rule to obtain a trained specified mixed model; subsequently, the specified mixed model is used as the market trend prediction model. Based on the above processing procedure, the application trains the initial mixed model based on the combination of the multi-modal sample data related to the target industry and the industry knowledge, which can efficiently and accurately construct the prediction model with market trend prediction function, improves the construction efficiency of the prediction model, and ensures the model effect of the obtained prediction model.

[0100] In some optional implementation manners of the embodiment, after step S205, the electronic device can further perform the following steps:

[0101] Obtain a real market trend result matched with the market trend prediction result.

[0102] In the embodiment, the real market trend result refers to the real market trend data corresponding to the market trend prediction result.

[0103] Based on a preset calculation strategy, the market trend prediction result and the real market trend result are compared and calculated to obtain corresponding deviation data.

[0104] In the embodiment, the calculation strategy can be a strategy measured by absolute error, relative error, etc. According to the selected calculation strategy, the market trend prediction result and the real market trend result are compared, the deviation between them is calculated, and the obtained calculation result is used as the corresponding deviation data.

[0105] Call a preset proximal strategy optimization algorithm.

[0106] In the embodiment, the proximal strategy optimization algorithm can also be referred to as a PPO algorithm.

[0107] Based on the proximal policy optimization algorithm, the prediction model is optimized using the deviation data to obtain a corresponding target prediction model.

[0108] In this embodiment, after the deviation is calculated, the hyperparameter combination of the prediction model (LSTM-CNN hybrid model) is adjusted according to the size and direction of the deviation by using the proximal policy optimization algorithm. Specifically, if the predicted value is higher than the true value and the deviation is large, the PPO algorithm adjusts the hyperparameters such as the number of hidden units of the LSTM layer and the size of the convolution kernel of the CNN layer, so that the model can more accurately reflect the market trend in subsequent prediction.

[0109] Based on the target prediction model, the prediction model is updated.

[0110] In this embodiment, the original prediction model can be replaced by using the optimized target prediction model to perform subsequent market trend prediction processing, thereby completing the model update processing of the prediction model.

[0111] The present application obtains the real market trend result matched with the market trend prediction result, then compares and calculates the market trend prediction result and the real market trend result based on a preset calculation strategy to obtain corresponding deviation data, then calls a preset proximal policy optimization algorithm, and uses the deviation data to optimize the prediction model based on the proximal policy optimization algorithm to obtain a corresponding target prediction model. Based on the above processing flow, the present application obtains the real market trend result matched with the market trend prediction result, and compares and calculates the market trend prediction result and the real market trend result based on the use of the calculation strategy to obtain the deviation data, and then uses the deviation data to optimize the prediction model based on the proximal policy optimization algorithm to obtain the target prediction model, so that the updated target prediction model can continuously optimize its parameters according to the real-time market trend, which is beneficial to improve the accuracy of market trend prediction and timely reflect the impact of market sudden changes on emotions and market trend.

[0112] In some optional implementation manners of the present embodiment, after step S205, the above-mentioned electronic device can further perform the following steps:

[0113] Obtain a preset market index threshold.

[0114] In this embodiment, the value of the above-mentioned market index threshold is not specifically limited and can be set according to actual business needs, for example, a standard deviation threshold such as twice the standard deviation can be set as the above-mentioned market index threshold.

[0115] determine whether the market trend prediction result exceeds the market index threshold.

[0116] In this embodiment, the fluctuation of the market index can be determined by comparing the market trend prediction result with the market index threshold.

[0117] If yes, corresponding alarm information is generated based on the market trend prediction result.

[0118] In this embodiment, if the generated market trend prediction result exceeds the market index threshold, the market trend prediction result is intelligently filled into a preset alarm information template to generate corresponding alarm information. The content of the alarm information template is not specifically limited and can be set according to actual alarm requirements. For example, the generated alarm information can include the market trend prediction result, market sentiment, and abnormal change of the market trend.

[0119] The alarm information is sent to a corresponding user based on a preset notification mode.

[0120] In this embodiment, the notification mode matched with the user can be determined according to the user's settings, for example, can include sending an email, an SMS, or an API push. Further, the generated alarm information is sent to the corresponding user according to the selected notification mode, so as to timely remind the user of the abnormal change of the market sentiment and the market trend, and enable the user to make a decision in time.

[0121] The application obtains a preset market index threshold, then determines whether the market trend prediction result exceeds the market index threshold, if yes, corresponding alarm information is generated based on the market trend prediction result, and subsequently, the alarm information is sent to a corresponding user based on a preset notification mode. Based on the above processing procedure, when the market trend prediction result exceeds the market index threshold, the application automatically generates alarm information based on the market trend prediction result, and sends the alarm information to the corresponding user based on the use of the notification mode, so as to timely remind the user when the market has an abnormal change, provide timely decision support for the user, enable the user to better grasp the market dynamics, make a reasonable investment decision, and improve the user's use experience.

[0122] In some optional implementation manners, the obtained user information seeks the user's consent and meets the relevant legal and relevant policy requirements.

[0123] In addition, the non-company software tools or components appearing in the embodiments of the application are only examples and do not represent actual use.

[0124] It should be understood that the size of the serial number of each step in the above embodiments does not mean the order of execution, and the execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the application.

[0125] It should be emphasized that, in order to further ensure the privacy and security of the above market trend prediction results, the above market trend prediction results can also be stored in a node of a block chain.

[0126] The blockchain referred to in the present application is a new application mode of distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm and other computer technologies. Blockchain, in essence, is a decentralized database, which is a series of data blocks associated using cryptographic methods, each data block containing information of a batch of network transactions, used to verify the validity (anti-fake) of the information and generate the next block. The blockchain can include a blockchain underlying platform, a platform product service layer, and an application service layer.

[0127] The embodiments of the present application can acquire and process related data based on artificial intelligence technology. Among them, artificial intelligence (Artificial Intelligence, AI) is to use digital computers or digital computer controlled machines to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. Theory, method, technology and application system.

[0128] The basic technology of artificial intelligence generally includes technologies such as sensors, special artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction system, mechatronics, etc. Artificial intelligence software technology mainly includes computer vision technology, robot technology, biometric technology, speech processing technology, natural language processing technology, and machine learning / deep learning, etc. Several major directions.

[0129] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by computer readable instructions instructing related hardware, which can be stored in a computer readable storage medium. The program can include the processes of the above-mentioned embodiments when executed, wherein the storage medium can be a non-volatile storage medium such as a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0130] It should be understood that although each step in the flowchart of the accompanying drawings is shown in sequence according to the direction of the arrow, these steps are not necessarily executed in sequence according to the direction of the arrow. Unless explicitly stated herein, the execution of these steps is not strictly limited in sequence, and they can be executed in other sequences. Moreover, at least some of the steps in the flowchart of the accompanying drawings can include multiple sub-steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence is not necessarily sequential, but can be alternately executed with at least part of other steps or sub-steps or stages of other steps.

[0131] Further referring to Figure 3 , as an implementation of the method shown in the above Figure 2 , the present application provides an embodiment of a data prediction device, which corresponds to the method embodiment shown in Figure 2 , and the device can be specifically applied to various electronic devices.

[0132] As shown in Figure 3 , the data prediction device 300 described in the embodiment includes a collection module 301, a filtering module 302, an extraction module 303, an analysis module 304, a prediction module 305, and a display module 306. Among them:

[0133] The collection module 301 is configured to collect initial multi-modal data corresponding to a target industry from a designated platform; wherein the initial multi-modal data includes text data, image data, and audio data;

[0134] The filtering module 302 is configured to perform noise filtering on the initial multi-modal data based on a preset multi-level verification strategy to obtain corresponding multi-modal data;

[0135] The extraction module 303 is configured to perform emotion feature extraction on the multi-modal data to obtain corresponding emotion feature data;

[0136] The analysis module 304 is configured to perform emotion analysis on the emotion feature data based on a preset hybrid neural network to obtain a corresponding emotion analysis result;

[0137] The prediction module 305 is configured to perform prediction processing on the emotion analysis result based on a preset prediction model to obtain a corresponding market trend prediction result;

[0138] The display module 306 is configured to perform display processing on the emotion analysis result and the market trend prediction result.

[0139] In the embodiment, the above-mentioned modules or units are respectively used to perform operations corresponding to the steps of the data prediction method of the foregoing embodiments, which will not be described here.

[0140] In some optional implementation manners of the embodiment, the multi-level verification strategy includes an account-level verification strategy, a content-level verification strategy, and a propagation-level verification strategy; and the filtering module 302 includes:

[0141] a first processing submodule, configured to perform noise filtering on the initial multi-modal data based on the account-level verification strategy to obtain corresponding first processing data;

[0142] a second processing submodule, configured to perform noise filtering on the first processing data based on the content-level verification strategy to obtain corresponding second processing data;

[0143] a third processing submodule, configured to perform noise filtering on the second processing data based on the propagation-level verification strategy to obtain corresponding third processing data;

[0144] a first determining submodule, configured to take the third processing data as the multi-modal data.

[0145] In the embodiment, the operations performed by the above modules or units correspond to the steps of the data prediction method of the foregoing embodiments one by one, and will not be described here again.

[0146] In some optional implementation manners of the embodiment, the hybrid neural network includes a graph convolution network and a time sequence convolution network; and the analysis module 304 includes:

[0147] a first analysis submodule, configured to perform emotion analysis on the emotion feature data based on the graph convolution network to obtain corresponding first emotion information;

[0148] a second analysis submodule, configured to perform emotion analysis on the emotion feature data based on the time sequence convolution network to obtain corresponding second emotion information;

[0149] a fusion submodule, configured to perform fusion processing on the first emotion information and the second emotion information to obtain corresponding fusion information;

[0150] a second determining submodule, configured to take the fusion information as the emotion analysis result.

[0151] In the embodiment, the operations performed by the above modules or units correspond to the steps of the data prediction method of the foregoing embodiments one by one, and will not be described here again.

[0152] In some optional implementation manners of the embodiment, the display module 306 includes:

[0153] a first calling submodule, configured to call a preset counterfactual reasoning module;

[0154] a derivation submodule configured to derive the market trend prediction result based on the counterfactual reasoning module to obtain corresponding causal derivation information;

[0155] a second calling submodule configured to call a preset visual interface;

[0156] a display submodule configured to display the sentiment analysis result, the market trend prediction result, and the causal derivation information in the visual interface based on a preset multi-dimensional display strategy.

[0157] In the embodiment, the above modules or units are respectively used for performing operations corresponding to steps of the data prediction method of the foregoing embodiments, and thus will not be described here again.

[0158] In some optional implementations of the embodiment, the data prediction apparatus further includes:

[0159] a first obtaining module configured to obtain pre-constructed multi-modal sample data corresponding to the target industry;

[0160] a construction module configured to obtain industry knowledge corresponding to the target industry, and construct a corresponding mapping rule based on the industry knowledge;

[0161] a first calling module configured to call a preset initial mixed model;

[0162] a training module configured to train the initial mixed model based on the multi-modal sample data and the mapping rule to obtain a trained specified mixed model;

[0163] a determination module configured to determine the specified mixed model as the market trend prediction model.

[0164] In the embodiment, the above modules or units are respectively used for performing operations corresponding to steps of the data prediction method of the foregoing embodiments, and thus will not be described here again.

[0165] In some optional implementations of the embodiment, the data prediction apparatus further includes:

[0166] a second obtaining module configured to obtain a real market trend result matched with the market trend prediction result;

[0167] a calculation module configured to perform comparison calculation on the market trend prediction result and the real market trend result based on a preset calculation strategy to obtain corresponding deviation data;

[0168] a second calling module configured to call a preset proximal policy optimization algorithm;

[0169] An optimization module is used to optimize the prediction model using the deviation data based on the near-end strategy optimization algorithm to obtain the corresponding target prediction model.

[0170] The update module is used to perform model update processing on the prediction model based on the target prediction model.

[0171] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the data prediction method in the aforementioned embodiments, and will not be repeated here.

[0172] In some optional implementations of this embodiment, the data prediction device further includes:

[0173] The third acquisition module is used to acquire preset market indicator thresholds;

[0174] The judgment module is used to determine whether the market prediction result exceeds the market indicator threshold;

[0175] The generation module is used to generate corresponding alarm information based on the market prediction results if the situation is as described.

[0176] The sending module is used to send the alarm information to the corresponding user based on a preset notification method.

[0177] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the data prediction method in the aforementioned embodiments, and will not be repeated here.

[0178] To address the aforementioned technical problems, embodiments of this application also provide a computer device. Please refer to [link / reference needed]. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0179] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected via a system bus. It should be noted that only the computer device 4 with components 41-43 is shown in the figure; however, it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively. Those skilled in the art will understand that the computer device described here is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0180] The computer device can be a desktop computer, a notebook computer, a palm computer, a cloud server, or the like. The computer device can interact with a user through a keyboard, a mouse, a remote controller, a touchpad, a voice control device, or the like.

[0181] The memory 41 can include at least one type of readable storage medium, such as a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., an SD or DX memory, or the like), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, or the like. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or a memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, or the like. Of course, the memory 41 can include both an internal storage unit and an external storage device of the computer device 4. In this embodiment, the memory 41 is generally used to store an operating system and various application software installed in the computer device 4, such as computer readable instructions of the data prediction method, or the like. In addition, the memory 41 can also be used to temporarily store various data that has been output or will be output.

[0182] The processor 42 can be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip in some embodiments. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to run computer readable instructions or process data stored in the memory 41, such as computer readable instructions of the data prediction method.

[0183] The network interface 43 can include a wireless network interface or a wired network interface, and is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0184] The present application also provides another embodiment, i.e., a computer readable storage medium storing computer readable instructions, which can be executed by at least one processor to make the at least one processor perform the steps of the data prediction method as described above.

[0185] Those skilled in the art can clearly understand the above-mentioned embodiment method can be realized by means of software and necessary general hardware platform, of course, also can be realized by hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application essentially or say the part of the prior art to make contributions can be embodied in the form of software products, the computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disc), including a number of instructions to make a terminal device (may be a mobile phone, computer, server, air conditioner, or network equipment, etc.) executes the method described in various embodiments of the present application.

[0186] Obviously, the above-described embodiments are only a part of the embodiments of the present application, rather than all the embodiments, the preferred embodiments of the present application are given in the drawings, but do not limit the patent scope of the present application. The present application can be realized in many different forms, and contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions recorded in the foregoing specific embodiments, or make equivalent replacement for part of the technical features. Any equivalent structure made by using the contents of the present application specification and drawings, directly or indirectly used in other related technical fields, are also within the scope of the present application patent protection.

Claims

1. A data prediction method characterized by, The method comprises the following steps: Collecting initial multi-modal data corresponding to a target industry from a designated platform, wherein the initial multi-modal data comprises text data, image data, and audio data; Filtering noise from the initial multi-modal data based on a preset multi-level verification strategy to obtain corresponding multi-modal data; Extracting emotional features from the multi-modal data to obtain corresponding emotional feature data; Performing emotional analysis on the emotional feature data based on a preset hybrid neural network to obtain corresponding emotional analysis results; Performing prediction processing on the emotional analysis results based on a preset prediction model to obtain corresponding market trend prediction results; Displaying the emotional analysis results and the market trend prediction results.

2. The data prediction method of claim 1, wherein, The multi-level verification strategy comprises an account-level verification strategy, a content-level verification strategy, and a propagation-level verification strategy. The step of filtering noise from the initial multi-modal data based on the preset multi-level verification strategy to obtain corresponding multi-modal data comprises: Filtering noise from the initial multi-modal data based on the account-level verification strategy to obtain corresponding first processed data; Filtering noise from the first processed data based on the content-level verification strategy to obtain corresponding second processed data; Filtering noise from the second processed data based on the propagation-level verification strategy to obtain corresponding third processed data; Using the third processed data as the multi-modal data.

3. The data prediction method of claim 1, wherein, The hybrid neural network comprises a graph convolution network and a time series convolution network. The step of performing emotional analysis on the emotional feature data based on the preset hybrid neural network to obtain corresponding emotional analysis results comprises: Performing emotional analysis on the emotional feature data based on the graph convolution network to obtain corresponding first emotional information; Performing emotional analysis on the emotional feature data based on the time series convolution network to obtain corresponding second emotional information; Fusing the first emotional information and the second emotional information to obtain corresponding fused information; Using the fused information as the emotional analysis results.

4. The data prediction method of claim 1, wherein, The step of displaying the emotional analysis results and the market trend prediction results comprises: Calling a preset counterfactual reasoning module; Performing deduction processing on the market trend prediction results based on the counterfactual reasoning module to obtain corresponding causal deduction information; Calling a preset visualization interface; Displaying the emotional analysis results, the market trend prediction results, and the causal deduction information in the visualization interface based on a preset multi-dimensional display strategy.

5. The data prediction method of claim 1, wherein, Before the step of performing prediction processing on the emotional analysis results based on the preset prediction model to obtain corresponding market trend prediction results, the method further comprises: Obtaining pre-constructed multi-modal sample data corresponding to the target industry; Obtaining industry knowledge corresponding to the target industry and constructing corresponding mapping rules based on the industry knowledge; Calling a preset initial hybrid model; Training the initial hybrid model based on the multi-modal sample data and the mapping rules to obtain a trained designated hybrid model; The designated mixed model is taken as the market trend prediction model.

6. The data prediction method of claim 1, wherein, After the step of performing prediction processing on the emotion analysis result based on the preset prediction model to obtain a corresponding market trend prediction result, the method further includes: obtaining a real market trend result matched with the market trend prediction result; performing comparison calculation on the market trend prediction result and the real market trend result based on a preset calculation strategy to obtain corresponding deviation data; calling a preset proximal strategy optimization algorithm; performing optimization processing on the prediction model based on the proximal strategy optimization algorithm using the deviation data to obtain a corresponding target prediction model; performing model update processing on the prediction model based on the target prediction model.

7. The data prediction method of claim 1, wherein, After the step of performing prediction processing on the emotion analysis result based on the preset prediction model to obtain a corresponding market trend prediction result, the method further includes: obtaining a preset market trend index threshold value; determining whether the market trend prediction result exceeds the market trend index threshold value; if yes, generating corresponding alarm information based on the market trend prediction result; sending the alarm information to a corresponding user based on a preset notification mode.

8. A data prediction apparatus, characterized by comprising: The method includes: a collection module configured to collect initial multi-modal data corresponding to a target industry from a designated platform, wherein the initial multi-modal data includes text data, image data, and audio data; a filtering module configured to perform noise filtering on the initial multi-modal data based on a preset multi-level verification strategy to obtain corresponding multi-modal data; an extraction module configured to perform emotion feature extraction on the multi-modal data to obtain corresponding emotion feature data; an analysis module configured to perform emotion analysis on the emotion feature data based on a preset mixed neural network to obtain corresponding emotion analysis result; a prediction module configured to perform prediction processing on the emotion analysis result based on a preset prediction model to obtain a corresponding market trend prediction result; a display module configured to perform display processing on the emotion analysis result and the market trend prediction result.

9. A computer device, comprising: The computer readable storage medium stores computer readable instructions, and the computer readable instructions are executed by the processor to realize the steps of the data prediction method in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer readable instructions, and the computer readable instructions are executed by the processor to realize the steps of the data prediction method in any one of claims 1 to 7.

Images