WO2024021354A1

WO2024021354A1 - Model training method, price prediction method, terminal device and storage medium

Info

Publication number: WO2024021354A1
Application number: PCT/CN2022/129587
Authority: WO
Inventors: 吴胤旭; 叶子; 陈会; 周亚雯; 姜青山
Original assignee: 中国科学院深圳先进技术研究院
Priority date: 2022-07-28
Filing date: 2022-11-03
Publication date: 2024-02-01
Also published as: CN115423499A

Abstract

Disclosed in the present application are a model training method based on a market price prediction model, a price prediction method based on a market price prediction model, and a terminal device and a computer storage medium. The model training method comprises: acquiring a data set; calculating a historical price technical index on the basis of stock market data of each time period; calculating a historical social media market sentiment index on the basis of social media data of each time period; associating and merging the historical price technical index and the historical social media market sentiment index in the same time period, and inputting same as features into a market price prediction model to be trained; and training the market price prediction model by using a prediction output of the market price prediction model to be trained, so as to obtain a final market price prediction model. By means of the model training method in the present application, the time granularity of prediction is reduced by means of the characteristic of continuous updating of social media, so as to obtain a capability of almost real-time prediction; and in view of a market technical index, a better prediction effect is achieved.

Description

Model training methods, price prediction methods, terminal equipment and storage media

Technical field

This application relates to the field of financial technology technology, and in particular to a model training method based on a market price prediction model, a price prediction method, a terminal device and a computer storage medium.

Background technique

Throughout the history of currency development, currency forms have evolved from physical currency and metal currency to convertible banknotes, non-convertible credit currency, and then to electronic currency and digital currency. Technological innovation not only drives social development, but also changes the form of currency. In the past thirty years, the popularization and development of Internet technology have greatly promoted the changes in payment and settlement methods, and promoted the development of e-commerce, e-government and digital economy. People's demand for paper money is getting smaller and smaller, and is replaced by electronic money and digital money. Currently, our country is focusing on developing a modern economic system and increasing efforts to promote the digital economy. Against this background, many scholars have focused their research on digital currency. In January 2017, the Central Bank of my country officially established the Digital Currency Research Institute. In April 2020, the legal digital currency independently designed and issued by the Central Bank was pilot tested in Suzhou, Shenzhen, Xiongan and Chengdu. As an "emerging thing", digital currency is receiving more and more attention. The gradual development of digital currency and digital economy will also change the pattern of the financial industry. Its virtualization and other features will put forward higher and faster requirements for the financial industry. Fintech will usher in more technological upgrades and updates, and research on how to use artificial intelligence The application of intelligence in the financial industry will also have more practical significance.

At the same time, with the development of financial technology, various market price prediction models based on machine learning and artificial intelligence have attracted more and more attention from all walks of life. Among them, models incorporating sentiment analysis are one of the mainstream research directions in this field. This model that combines text analysis uses text and price data, so how to extract features from text data becomes a major challenge. With the rapid development of Internet finance and securities markets, the number of new investors in my country has hit new highs. Social media such as stock bars, major financial forums, and Weibo have become important media for stock market investors to share information and make investment references. Market inefficiency theory explains that stock prices cannot fully reflect stock value, that is, there is a certain correlation between investor sentiment and stock price trends. Based on the sentiment analysis of stock comments by investors in the stock bar and the time series analysis of stock price-related information, the rise and fall of stocks can be predicted to a certain extent, thereby providing investment advice to small and medium-sized investors. At the same time, the conclusions of sentiment analysis of stock comments are also It can provide analytical basis for institutional investors and stock analysts. The predictive factors used in traditional quantitative investment are generally stock-related data or macro- and micro-economic data, and do not model the inefficiency of the market.

Contents of the invention

This application provides a model training method, a price prediction method, a terminal device and a computer storage medium based on a market price prediction model.

One technical solution adopted by this application is to provide a model training method based on a market price prediction model. The model training method includes:

Obtain a data set, wherein the data set includes stock market data and social media data for several time periods;

Calculate historical price technical indicators based on stock market data for each time period;

Based on the social media data of each time period, calculate the historical social media market sentiment index;

Correlate and merge historical price technical indicators and historical social media market sentiment index in the same time period, and input them as features into the market price prediction model to be trained;

The market price prediction model is trained using the prediction output of the market price prediction model to be trained to obtain the final market price prediction model.

Wherein, the stock market data includes opening price, highest price, lowest price, closing price and/or trading volume.

Wherein, the historical price technical indicators include moving averages of similarity and difference, simple moving averages, relative strength indicators and/or fund flow indicators;

Based on the stock market data of each time period, the historical price technical indicators are calculated, including:

Calculate the moving average of similarity and difference based on the opening price of each time period;

and/or, calculating said simple moving average based on the highest price in each time period;

and/or, calculate the relative strength indicator based on the lowest price in each time period;

and/or, calculating the money flow indicator based on the closing price for each time period.

Among them, the historical social media market sentiment index is calculated based on the social media data of each time period, including:

Obtain the comment text of social media data in each time period and obtain the sentiment score of each comment text;

Based on the sentiment score of each comment text, comment texts lower than the preset threshold are defined as negative comment texts, and comment texts higher than the preset threshold are defined as positive comment texts;

Based on all positive comment texts and negative comment texts of the social media data in each time period, the gradient sentiment divergence index of the social media data is calculated.

Wherein, after calculating the gradient sentiment divergence index of the social media data, the model training method further includes:

Obtain the difference between each comment text of the social media data and the gradient sentiment deviation index;

The gradient bull market sentiment index of the social media data is calculated using the difference corresponding to all comment texts of the social media data and the number of all comment texts.

Among them, the correlation and merging of historical price technical indicators and historical social media market sentiment indexes in the same time period include:

Correlate and merge the opening price in the same time period with the moving average of similarity and difference and the gradient sentiment divergence index;

The highest price in the same time period is correlated and merged with the simple moving average and the gradient bull sentiment index.

Among them, the obtained data set includes:

Use crawler technology or the supplier's programming interface to collect stock market data;

and/or, use crawler technology or social media programming interfaces to collect social media data.

Another technical solution adopted by this application is to provide a price prediction method based on a market price prediction model. The price prediction method includes:

Get the stock market data for the current time period, and get all social comment data for the current time period;

Based on the stock market data, calculate price technical indicators for the current time period;

Calculate the social media market sentiment index of the current time period based on all social comment data;

Correlate and merge the price technical indicators and social media market sentiment index of the current time period, and input them as features into the pre-trained market price prediction model;

Based on the output of the market price prediction model, obtain the predicted price after the current time period;

Wherein, the market price prediction model is trained by the above-mentioned model training method.

Another technical solution adopted by this application is to provide a terminal device, which includes a memory and a processor coupled to the memory;

Wherein, the memory is used to store program data, and the processor is used to execute the program data to implement the above-mentioned model training method and/or the above-mentioned price prediction method.

Another technical solution adopted by this application is to provide a computer storage medium. The computer storage medium is used to store program data. When the program data is executed by a computer, it is used to implement the above-mentioned model training method and/or the above-mentioned price prediction method.

The beneficial effects of this application are: the terminal device obtains the data set; based on the stock market data of each time period, calculates historical price technical indicators; based on the social media data of each time period, calculates the historical social media market sentiment index; the same time period The historical price technical indicators and the historical social media market sentiment index are associated and merged, and are input as features to the market price prediction model to be trained; the market price prediction model is trained using the prediction output of the market price prediction model to be trained, and the final market price prediction model. The model training method of this application uses the continuous update characteristics of social media to reduce the time granularity of prediction and obtain close to real-time prediction capabilities. It also achieves better prediction results by combining market technical indicators.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without exerting creative efforts.

Figure 1 is a schematic flow chart of an embodiment of a model training method based on a market price prediction model provided by this application;

Figure 2 is a schematic diagram of the overall flow of the model training method and its price prediction method based on the market price prediction model provided by this application;

Figure 3 is a schematic diagram of the closing price predicted by the market price prediction model provided by this application;

Figure 4 is a schematic framework diagram of an embodiment of the data set provided by this application;

Figure 5 is a schematic flow chart of an embodiment of a price prediction method based on a market price prediction model provided by this application;

Figure 6 is a schematic structural diagram of an embodiment of a terminal device provided by this application;

Figure 7 is a schematic structural diagram of an embodiment of a computer storage medium provided by this application.

Detailed ways

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of this application.

In recent years, with the rapid development of Internet technology and the widespread popularity of social networks, a large amount of stock-related information has been spread on the Internet. The scale of this real-time information is huge, and it contains important information related to the stock market. Stock price trend prediction based on Internet and social network information has gradually become a mainstream research direction. At the same time, traditional stock price trend analysis and prediction methods have not taken into account the scale and timeliness of data, and have been unable to adapt to the massive data in the network environment. Stock market analysis and forecasting requirements at scale.

Based on the above research ideas, research from universities in the United States and the United States found that the rise and fall of individual stocks can be predicted by analyzing a large number of Weibo messages posted on the social network Twitter. Pace University in the United States tracked the stock prices of three companies: Starbucks, Coca-Cola and Nike, and concluded that the daily trend of a company's stock price can be predicted by a company's popularity on social media.

Researchers at the Technical University of Munich in Germany used information contained in Twitter messages to predict individual stock trends. The university previously conducted a study in which sentiment analysis methods were used to analyze Twitter messages related to certain stocks and whether those messages contained "bullish," "bearish" or "hold" messages.

This type of prediction technology makes use of market price data, and also obtains a large number of user comment texts through social media for sentiment analysis of big data to conduct market predictions. A common model is a deep neural network.

Based on the above foundation, this application aims to propose a neural network prediction method combined with social media sentiment analysis on the issue of financial market prediction. This method can effectively utilize data sets and improve prediction accuracy.

Please refer to Figures 1 and 2 for details. Figure 1 is a flow diagram of an embodiment of a model training method based on a market price prediction model provided by this application. Figure 2 is a model training method based on a market price prediction model provided by this application and its Schematic diagram of the overall process of price forecasting method.

As shown in Figure 1, the model training method based on the market price prediction model in the embodiment of this application specifically includes the following steps:

Step S11: Obtain a data set, where the data set includes stock market data and social media data for several time periods.

In the embodiment of this application, the data sources of stock market data in the data set include but are not limited to various securities firms and trading markets. The collection method can use crawler technology or collect historical price data through the supplier's programming interface.

Among them, the stock market data must be K-line data with specified time granularity, and the fields must include at least: opening price, closing price, highest price, lowest price, and trading volume. For example, if you specify a time granularity of 30 minutes, then there will be a record every 30 minutes from the opening of the market, recording the opening price, closing price, highest price, lowest price, and trading volume within that time period.

Therefore, the stock market data needs to contain the following fields:

The data sources of social media data in the data set include but are not limited to various social media, such as Weibo, Tieba, Twitter, user comment data in trading markets, etc. The collection method can use crawler technology or social media programming interfaces to collect historical social media data . Among them, historical social media data needs to include at least two fields: publishing time and published text content.

Step S12: Calculate historical price technical indicators based on the stock market data in each time period.

In the embodiment of this application, the historical price technical indicators used include but are not limited to: one or more of the moving average of similarity and difference, simple moving average, relative strength indicator and fund flow indicator.

The terminal device can calculate the moving average of similarities and differences based on the opening price of each time period; calculate the simple moving average based on the highest price of each time period; calculate the relative moving average based on the lowest price of each time period. Strength indicator; based on the closing price of each time period, the money flow indicator is calculated.

Specifically, the calculation method of the above historical price technical indicators is as follows:

1. MACD (Moving Average Convergence/Divergence) is the Moving Average Convergence/Divergence, which is developed from the double exponential moving average. It is obtained by subtracting the slow exponential moving average (EMA26) from the fast exponential moving average (EMA12). Fast line DIF, and then use 2× (9-day weighted moving average DEA of fast line DIF-DIF) to get the MACD column. The meaning of MACD is basically the same as that of the double moving average, that is, the current long and short status and the possible development trend of the stock price are represented by the dispersion and aggregation of the fast and slow moving averages, but it is more convenient to read. Changes in MACD represent changes in market trends, and MACD at different K-line levels represents the buying and selling trend in the current level cycle. The formula is:

MACD=EMA _(12-period) -EMA _(26-period)

Among them, EMA is the exponential average indicator. EMA (Exponential Moving Average) is an exponential moving average, also called EXPMA indicator. It is also a trend indicator. The exponential moving average is a moving average weighted by exponential decreasing.

2. SMA is a simple moving average, and the simple moving average is an arithmetic moving average, which is a simple and universal moving average. The formula is:

Wherein, p _n is the nth value, n is the moving window, and in the embodiment of the present application, n=30.

3.RSI is the relative strength indicator, which is the most famous oscillator in the futures market and stock market. The principle is to estimate the strength of the market trend by calculating the rise and fall of stock prices, and predict the continuation or reversal of the trend based on this. In fact, it shows the upward fluctuation of the stock price as a percentage of the total fluctuation. If the value is large, it means that the market is in a strong state. If the value is small, it means that the market is in a weak state. The formula is:

in,

In the embodiment of this application, n=14.

4MFI (Money Flow Index) is a fund flow indicator, which is a combination of the relative strength index (RSI) and the popularity index (OBV). The MFI indicator can be used to measure the momentum of trading volume and investment interest, and changes in trading volume provide clues to future changes in stock prices, so the MFI indicator can help determine the trend of stock price changes. The formula is:

in,

In the embodiment of this application, n=14.

At this point, the data in the current dataset can contain the following fields:

Step S13: Calculate the historical social media market sentiment index based on the social media data in each time period.

In the embodiment of this application, the terminal device obtains the comment text of the social media data in each time period, and obtains the emotion score of each comment text; based on the emotion score of each comment text, comments that are lower than the preset threshold are Text is defined as negative comment text, and comment text higher than the preset threshold is defined as positive comment text; based on all positive comment texts and negative comment texts of social media data in each time period, the social media Gradient Sentiment Divergence Index of the data.

Further, the terminal device can also obtain the difference between each comment text of the social media data and the gradient sentiment deviation index; use the difference corresponding to all comment texts of the social media data, and the number of all comment texts , calculate the gradient bull sentiment index for the social media data.

Specifically, the specific process of the terminal device calculating the historical social media market sentiment index is as follows:

The terminal device performs emotional scoring on the comment text of the social media data in each time period. Tools used for emotional scoring include but are not limited to VADER, etc. The terminal device performs an emotion score on each comment text and can obtain a value between -1 and 1, which represents the emotion of the comment text.

The terminal device may also define the comment text with the sentiment score between -1 and 0 as negative comment text, and the comment text with the sentiment score between 0 and 1 as positive comment text.

Specifically, the terminal device uses the defined positive comment text and negative comment text to calculate the Gradient Bullish Sentiment Index (Small Granular Sentiment Bullish Index, SGSDI) and the Gradient Sentiment Divergence Index (Small Granular Sentimental Divergence Index, SGSBI).

Among them, the calculation formula of the gradient sentiment divergence index is:

in,

is the sum of the sentiment index of all positive comments in the t time period,

It is the sum of the absolute values of the sentiment index of all negative comments in the t time period.

Among them, the calculation formula of the gradient bull market sentiment index is:

Among them, D(t) is the sentiment index set of all comments in the t time period.

Step S14: Correlate and merge the historical price technical indicators and the historical social media market sentiment index in the same time period, and input them as features into the market price prediction model to be trained.

In the embodiment of this application, the terminal device associates and merges the price data, technical indicators and emotional indicators of the same time period according to the time granularity area corresponding to the above data. At this time, the data set contains the following attributes:

Step S15: Use the prediction output of the market price prediction model to be trained to train the market price prediction model to obtain the final market price prediction model.

In the embodiment of the present application, the terminal device inputs the above-mentioned combined data set as a feature and adds a prediction target column. In the market price prediction model of the embodiment of the present application, the prediction target column is the closing price of the next time granularity, Please refer to Figure 3 for details. Figure 3 is a schematic diagram of the closing price predicted by the market price prediction model provided by this application.

As shown in Figure 3, the terminal device can input the data set from September 24, 2017 to September 27, 2017 into the market price prediction model. The time granularity is set to 1 day, and the prediction result output by the market price prediction model is 2017. The closing price on September 29, 2019. The terminal device can also input the data set from 00:00:00 on September 24, 2017 to 02:00:00 on September 24, 2017 into the market price prediction model. The time granularity is set to 30 minutes, and the market price prediction model outputs The prediction result is the closing price at 02:30:00 on September 24, 2017.

Among them, regarding the data set in step S11, the terminal device can divide the data set into a training set and a test set. Usually, 80% is used as the training set and 20% is used as the test set. For example, as shown in Figure 4 below, Figure 4 is the application A schematic diagram of the framework of an embodiment of the data set is provided.

As shown in Figure 4, the data set of the embodiment of this application includes data collected from 00:00:00 on September 24, 2017 to 23:59:59 on November 30, 2020. The terminal device will be on September 24, 2017. The data from 00:00:00 on April 11, 2020 to 23:59:59 on April 11, 2020 is used as the training set, and the data from 00:00:00 on April 12, 2020 to 23:59:59 on November 30, 2020 is used as the test set. set.

The training set in the embodiment of this application is only used to train the model. Generally, a deep neural network model is used, such as a long-short-term memory model (Long-Short Term Memory, LSTM) and other models. The training parameters need to be adjusted according to the data situation. In addition, it can also be used For more complex integrated models, the training cycle may be longer, but the model effect is generally slightly higher than that of a single deep neural network model. The test set is used to test the trained market price prediction model.

Among them, the market price prediction model can use various regressors, including but not limited to linear regression, deep neural networks, and various integrated regressors. For the evaluation indicators of the market price prediction model, you can choose the mean absolute error (Mean Absolute Error, MAE), mean square error (Mean Square Error, MSE), root mean square error (Root Mean Square Error, RMSE), R ² (R squared , Coefficient of determination), Mean Absolute Percentage Error (MAPE), etc. When each of the above evaluation indicators reaches an acceptable range, the market price prediction model is saved.

In order to verify the effectiveness and advancement of the technical route proposed by this patent, multiple extensive experiments were conducted on the real data sets of Twitter and Bitcoin to evaluate the performance of the proposed method.

The three evaluation indicators of mean absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE) are used to evaluate the model effect and compared with existing models. The experimental results consistently show that using this method The patented method can predict prices more accurately and can be widely used to predict the prices of financial products, and has very broad application prospects.

The following Table 1 and Table 2 show the prediction error statistics of the two models at different granularities at two times. LSTM is the Long-Short Term Memory model (Long-Short Term Memory) and GRU is the Gate Recurrent Unit model (Gate Recurrent Unit). The data for this simulation uses Bitcoin price data and #Bitcoin topic tweets on Twitter from September 24, 2017 to November 30, 2020. The following results were obtained using 80% training set and 20% test set.

Table 1 Model prediction performance results table at 30-minute time granularity

Table 2 Model prediction performance results table at 1-day time granularity

In the embodiment of this application, the continuous updating characteristics of social media are used to replace news data, so that the time granularity of prediction can be further reduced, and the ability of near-real-time prediction is obtained; this application provides a new market sentiment within a time period Compared with the current simple statistical method, the index calculation method can better reflect the specific market sentiment and improve the accuracy of the model.

Based on the model training method based on the market price prediction model in the above embodiment, this application further proposes a price prediction method based on the market price prediction model. For details, please refer to Figure 5. Figure 5 is a model based on the market price prediction model provided by this application. A flowchart of an embodiment of a price prediction method.

As shown in Figure 5, the price prediction method in the embodiment of this application specifically includes the following steps:

Step S21: Obtain the stock market data of the current time period and obtain all social comment data of the current time period.

Step S22: Calculate price technical indicators of the current time period based on the stock market data.

Step S23: Calculate the social media market sentiment index of the current time period based on all social comment data.

Step S24: Correlate and merge the price technical indicators and the social media market sentiment index of the current time period, and input them as features into the pre-trained market price prediction model.

Step S25: Based on the output of the market price prediction model, obtain the predicted price after the current time period.

Please continue to refer to Figure 6. Figure 5 is a schematic structural diagram of an embodiment of a terminal device provided by this application. The terminal device 500 in the embodiment of the present application includes a processor 51, a memory 52, an input and output device 53, and a bus 54.

The processor 51, the memory 52, and the input and output device 53 are respectively connected to the bus 54. The memory 52 stores program data. The processor 51 is used to execute the program data to implement the model training method and/or price prediction described in the above embodiment. method.

In the embodiment of this application, the processor 51 may also be called a CPU (Central Processing Unit). The processor 51 may be an integrated circuit chip with signal processing capabilities. The processor 51 can also be a general-purpose processor, a digital signal processor (DSP, Digital Signal Process), an application specific integrated circuit (ASIC, Application Specific Integrated Circuit), a field programmable gate array (FPGA, Field Programmable Gate Array) or other available Programmed logic devices, discrete gate or transistor logic devices, discrete hardware components. The general processor may be a microprocessor or the processor 51 may be any conventional processor or the like.

This application also provides a computer storage medium. Please continue to refer to Figure 7. Figure 7 is a schematic structural diagram of an embodiment of the computer storage medium provided by this application. The computer storage medium 600 stores program data 61. The program data 61 is in When executed by the processor, it is used to implement the model training method and price prediction method of the above embodiment.

When the embodiments of the present application are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or contributes to the existing technology, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium , including several instructions to cause a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the method described in each embodiment of the application. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program code. .

The above descriptions are only embodiments of the present application and are not intended to limit the patent scope of the present application. Equivalent structures or equivalent process transformations may be made using the contents of the description and drawings of the present application, or directly or indirectly applied to other related technologies. fields are equally included in the scope of patent protection of this application.

Claims

A model training method based on a market price prediction model, characterized in that the model training method includes:

Obtain a data set, wherein the data set includes stock market data and social media data for several time periods;

Calculate historical price technical indicators based on stock market data for each time period;

Based on the social media data of each time period, calculate the historical social media market sentiment index;

Correlate and merge historical price technical indicators and historical social media market sentiment index in the same time period, and input them as features into the market price prediction model to be trained;

The market price prediction model is trained using the prediction output of the market price prediction model to be trained to obtain the final market price prediction model.
The model training method according to claim 1, characterized in that:

The stock market data includes opening price, high price, low price, closing price and/or trading volume.
The model training method according to claim 2, characterized in that:

The historical price technical indicators include moving averages of similarity and divergence, simple moving averages, relative strength indicators and/or fund flow indicators;

Based on the stock market data of each time period, the historical price technical indicators are calculated, including:

Calculate the moving average of similarity and difference based on the opening price of each time period;

and/or, calculating said simple moving average based on the highest price in each time period;

and/or, calculate the relative strength indicator based on the lowest price in each time period;

and/or, calculating the money flow indicator based on the closing price for each time period.
The model training method according to claim 1, characterized in that:

The historical social media market sentiment index is calculated based on the social media data of each time period, including:

Obtain the comment text of social media data in each time period and obtain the sentiment score of each comment text;

Based on the sentiment score of each comment text, comment texts lower than the preset threshold are defined as negative comment texts, and comment texts higher than the preset threshold are defined as positive comment texts;

Based on all positive comment texts and negative comment texts of the social media data in each time period, the gradient sentiment divergence index of the social media data is calculated.
The model training method according to claim 4, characterized in that:

After calculating the gradient sentiment divergence index of the social media data, the model training method further includes:

Obtain the difference between each comment text of the social media data and the gradient sentiment deviation index;

The gradient bull market sentiment index of the social media data is calculated using the difference corresponding to all comment texts of the social media data and the number of all comment texts.
The model training method according to claim 5, characterized in that:

The above-mentioned correlation and merging of historical price technical indicators and historical social media market sentiment index in the same time period includes:

Correlate and merge the opening price in the same time period with the moving average of similarity and difference and the gradient sentiment divergence index;

The highest price in the same time period is correlated and merged with the simple moving average and the gradient bull sentiment index.
The model training method according to claim 1, characterized in that:

The obtained data set includes:

Use crawler technology or the supplier's programming interface to collect stock market data;

and/or, use crawler technology or social media programming interfaces to collect social media data.
A price prediction method based on a market price prediction model, characterized in that the price prediction method includes:

Get the stock market data for the current time period, and get all social comment data for the current time period;

Based on the stock market data, calculate price technical indicators for the current time period;

Calculate the social media market sentiment index of the current time period based on all social comment data;

Correlate and merge the price technical indicators and social media market sentiment index of the current time period, and input them as features into the pre-trained market price prediction model;

Based on the output of the market price prediction model, obtain the predicted price after the current time period;

Wherein, the market price prediction model is trained by the model training method described in any one of claims 1 to 7.
A terminal device, characterized in that the terminal device includes a memory and a processor coupled to the memory;

Wherein, the memory is used to store program data, and the processor is used to execute the program data to implement the model training method according to any one of claims 1 to 7 and/or the price prediction method according to claim 8. .
A computer storage medium, characterized in that the computer storage medium is used to store program data, and when the program data is executed by a computer, it is used to implement the model training method as described in any one of claims 1 to 7. /or the price prediction method according to claim 8.