System for analyzing future futures market through combination of historical futures linear graph and big data
By combining historical futures linear graphs and big data analysis, the LSTM and Transformer model and reinforcement learning algorithm are used to solve the problem that traditional prediction methods are difficult to capture market dynamic changes and identify complex nonlinear relationships, and high-precision and real-time futures market forecasting and decision-making support are achieved.
Patent Information
- Application Number
- CN202510372055.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-27
- Publication Date
- 2025-06-27
AI Technical Summary
Traditional futures market price prediction methods are difficult to capture market dynamic changes in real time, and cannot fully explore complex nonlinear relationships, resulting in low prediction accuracy.
Through a system that combines historical futures linear charts with big data to analyze future futures markets, multi-source data is integrated, LSTM and Transformer model are combined, reinforcement learning algorithms and model fusion technology are introduced to generate trading signals and provide interpretability analysis.
It realizes high-precision and real-time prediction of futures markets, provides explainable decision-making support, and improves investors' decision-making efficiency and return level in the futures market.
Smart Images
Figure FT_1 
Figure FT_2
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of financial data analysis, and particularly to a system for analyzing future futures market conditions by combining historical futures line charts with big data analysis. Background Art
[0002] In the futures market, accurately predicting price trends is of great significance to investors and trading institutions. Traditional prediction methods mostly rely on historical data for statistical modeling, such as time series models like ARIMA and machine learning models like support vector machines. However, the market is complex and ever-changing, and these traditional methods have significant drawbacks. On the one hand, it is difficult to capture market dynamic changes in real time, showing obvious lag; on the other hand, the price fluctuations in the futures market present complex non-linear relationships, and traditional models cannot fully explore and understand these complex patterns, resulting in the prediction accuracy being difficult to meet the actual needs.
[0003] Therefore, a system for analyzing future futures market conditions by combining historical futures line charts with big data analysis is needed to achieve high-precision and real-time prediction of futures market conditions, provide interpretable decision-making support for investors, and effectively improve the decision-making efficiency and profit level of investors in the futures market. Summary of the Invention
[0004] In view of the deficiencies of the prior art, the present invention provides a system for analyzing future futures market conditions by combining historical futures line charts with big data analysis, which solves the problems raised in the above background art.
[0005] Technical Solution: To solve the above technical problems, according to one aspect of the present invention, more specifically, a system for analyzing future futures market conditions by combining historical futures line charts with big data analysis includes a data acquisition layer, a data preprocessing layer, a feature engineering layer, a prediction model layer, and a decision support layer;
[0006] The data acquisition layer is used to collect futures historical prices, trading volumes, open interest, news reports, investor discussion information, and macroeconomic indicator data from futures exchanges, news media, social platforms, government economic databases, and authoritative financial institution websites;
[0007] The data preprocessing layer includes a data cleaning module, a data conversion module, and a data normalization module;
[0008] The feature engineering layer includes a technical indicator feature extraction module, an emotion feature extraction module, a macroeconomic feature integration module, and a composite feature construction module;
[0009] The prediction model layer includes an LSTM model construction module, a Transformer model integration module, a reinforcement learning optimization module, and a model fusion and tuning module;
[0010] The decision support layer includes a trading signal generation module, an interpretability analysis and display module, and a risk assessment and prompt module.
[0011] Furthermore, in the data acquisition layer, futures exchange data is collected in real time through an API interface, news media and social platform data are collected regularly through web crawlers, and macroeconomic data is obtained from relevant websites regularly.
[0012] Furthermore, in the data preprocessing layer, the data cleaning module is used to identify and correct outliers and fill in missing values. It uses the IQR algorithm to identify outliers and the time series interpolation method to fill in missing values.
[0013] The data conversion module is used to convert unstructured data into structured data.
[0014] The data normalization module is used to eliminate the influence of different dimensions. For numerical data, the Z-score standardization or Min-Max normalization method is adopted.
[0015] Furthermore, in the feature engineering layer, the technical indicator feature extraction module calculates technical analysis indicators such as the Relative Strength Index (RSI), Moving Average Convergence Divergence (MACD), and Bollinger Bands.
[0016] The sentiment feature extraction module uses the BERT pre-trained language model for sentiment analysis.
[0017] The macroeconomic feature integration module correlates and integrates macroeconomic indicator data and futures market data to construct a relationship model between macroeconomic factors and the futures market.
[0018] The composite feature construction module constructs a composite feature vector and performs dimensionality reduction processing. The dimensionality of the features is reduced through the principal component analysis (PCA) dimensionality reduction technique.
[0019] Furthermore, in the prediction model layer, the LSTM model construction module is used to handle the long-term dependence relationship of time series.
[0020] The Transformer model integration module is combined with the LSTM model to capture the data correlation at different time points.
[0021] The reinforcement learning optimization module adopts the Proximal Policy Optimization (PPO) reinforcement learning algorithm to adjust the parameters according to the real-time feedback of the market.
[0022] The model fusion and tuning module adopts the Stacking and Bagging model fusion techniques and uses methods such as cross-validation and grid search to tune the model parameters.
[0023] Further, in the decision support layer, the trading signal generation module generates trading signals according to the prediction results in combination with preset trading strategies. The trading signals include buy, sell, or hold signals;
[0024] The interpretability analysis and display module uses SHAP value analysis and LIME technology to display the model decision-making process and presents the contribution degree of each feature to the prediction result to investors;
[0025] The risk assessment and prompt module uses the VaR value-at-risk model to assess investment risks, and assesses and prompts investment risks based on the prediction results and market fluctuations.
[0026] The beneficial effects of the system for predicting future futures market conditions by combining historical futures line charts with big data analysis in the present invention are as follows:
[0027] (1) By integrating multi-source data such as futures trading, news, social platforms, and macroeconomics, the present invention comprehensively reflects the market conditions, avoids single-data deviation, ensures the integrity and reliability of the data, and provides rich information for prediction.
[0028] (2) By combining LSTM and Transformer, the present invention can not only capture the long-term dependence of time series but also utilize the self-attention mechanism to grasp the correlations at different time points, effectively identify complex market patterns, and accurately predict market trends;
[0029] By introducing reinforcement learning algorithms such as PPO and adjusting the model parameters according to real-time market feedback, the prediction strategy is dynamically optimized, enabling the system to quickly adapt to market changes and continuously improve prediction performance;
[0030] By using techniques such as Stacking to fuse different models, giving full play to their respective advantages, and then tuning parameters through cross-validation and other methods, the model is ensured to be in the best state, improving prediction accuracy and stability. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] The present invention will be further described in detail below with reference to the drawings and specific implementation methods. It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments can be combined with each other.
[0032] Figure 1 is the flow chart of the present invention;
[0033] Figure 2 is the schematic diagram of the specific framework of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0034] The present invention will be described in detail below with reference to the drawings and in combination with embodiments. It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments can be combined with each other.
[0035] To make the technical solution of the present invention clearer, the following further describes the present invention in detail with reference to the accompanying drawings and specific embodiments.
[0036] Refer to Figure 1 - Figure 2 , a system for analyzing future futures market conditions by combining historical futures line charts with big data, including a data acquisition layer, a data preprocessing layer, a feature engineering layer, a prediction model layer, and a decision support layer;
[0037] Data acquisition layer: With the help of the exchange API interface, regularly and accurately collect core data such as historical futures prices (including opening price, highest price, lowest price, closing price), trading volume, and open interest, ensuring the integrity and accuracy of the data and laying a solid foundation for subsequent analysis. Use web crawler technology to comprehensively collect news reports related to the futures market from mainstream financial news websites and industry information platforms, and deeply extract key information and sentiment tendencies of the news with the help of natural language processing technology. For financial-related social groups, forums, etc., collect investors' discussion information and opinion expressions, and accurately mine market sentiment data through sentiment analysis algorithms. Obtain macroeconomic indicator data such as interest rates, exchange rates, GDP, and CPI from government economic databases and authoritative financial institution websites for in-depth analysis of the impact of the macroeconomic environment on the futures market.
[0038] The data preprocessing layer includes data cleaning: Use the IQR (Interquartile Range) algorithm to accurately identify and correct outliers in data such as prices and trading volumes to ensure the reliability of the data; for missing historical futures data, use time series interpolation methods (such as linear interpolation, cubic spline interpolation) to fill in and maintain data continuity.
[0039] Data conversion: Skillfully convert unstructured news texts and social platform data into structured data, such as converting text into an emotion index through sentiment analysis for subsequent in-depth analysis.
[0040] Data normalization: Use the Z-score standardization or Min-Max normalization method to standardize numerical data (such as prices, trading volumes, macroeconomic indicators) to eliminate the adverse effects of different dimensions on model training.
[0041] The feature engineering layer includes technical indicator feature extraction: Calculate common and effective technical analysis indicators, such as the Relative Strength Index (RSI), Moving Average Convergence Divergence (MACD), Bollinger Bands, etc., to provide a strong technical analysis basis for the model. Emotion feature extraction: Use pre-trained language models (such as BERT) to conduct in-depth sentiment analysis on news texts and social platform data to generate emotion indexes that can accurately reflect market sentiment. A positive emotion index reflects market optimism, and a negative emotion index reflects pessimism.
[0042] Integration of Macroeconomic Features: Deeply associate and integrate the collected macroeconomic indicator data with the futures market data. For example, analyze the impact of interest rate changes on futures prices and construct a close relationship model between macroeconomic factors and the futures market.
[0043] Construction of Composite Features: Organically integrate price features, technical indicator features, sentiment features, macroeconomic features, etc. to form a composite feature vector. And through dimensionality reduction techniques such as principal component analysis (PCA), while retaining the main information, significantly reduce the feature dimension and improve the model training efficiency.
[0044] The prediction model layer includes the construction of the LSTM model: The long short-term memory network (LSTM) can effectively handle the long-term dependence relationships in time series data. Construct an LSTM model and orderly input the historical data of the futures market that has undergone preprocessing and feature engineering into the model in chronological order, allowing the model to fully learn the long-term patterns of price movements.
[0045] Integration of the Transformer Model: The Transformer model has excellent parallel computing capabilities and self-attention mechanisms when dealing with long sequence data. Skillfully combine the Transformer model with the LSTM model, and use the self-attention mechanism of the Transformer to accurately capture the correlations between data at different time points, further significantly enhancing the model's ability to identify complex market patterns.
[0046] Optimization by Reinforcement Learning: Introduce reinforcement learning algorithms such as proximal policy optimization (PPO) to dynamically adjust the parameters of the prediction model according to the real-time feedback of the market. Carefully compare the prediction results with the actual market conditions, and guide the model to continuously optimize the prediction strategy through the reward mechanism to highly adapt to the dynamic changes of the market. Model Fusion and Tuning: Adopt model fusion techniques such as Stacking and Bagging to organically combine multiple different models (such as traditional machine learning models and deep learning models), integrate the advantages of each model, and improve the prediction accuracy and stability. Use methods such as cross-validation and grid search to finely tune the model parameters to ensure that the model is in the best performance state.
[0047] The decision support layer includes
[0048] Generation of Trading Signals: Based on the prediction results of the futures price trends output by the prediction model, combined with preset trading strategies (such as breakout strategies, mean reversion strategies, etc.), generate clear and guiding trading signals, such as buy, sell, or hold signals, to provide specific trading guidance for investors.
[0049] Interpretability analysis display: Use SHAP (SHapley Additive exPlanations) value analysis, LIME (Local Interpretable Model-agnostic Explanations) technology, etc. to visually display the model decision-making process. Clearly present to investors the contribution degree of each feature to the prediction result, help investors deeply understand the model decision-making logic, and enhance their trust in the system prediction result. Risk assessment and prompt: Based on the prediction result and market volatility, use a risk assessment model (such as VaR value-at-risk model) to scientifically assess the investment risk. When the risk exceeds the preset threshold, promptly issue a risk prompt to investors to assist them in doing a good job in risk management.
[0050] The above-described embodiments only express several implementation manners of the present invention. The description is relatively specific and detailed, but it cannot be construed as a limitation on the scope of the patent of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the appended claims.
Claims
1. A system for analyzing future futures market conditions by combining historical futures linear graphs with big data, including a data collection layer, a data preprocessing layer, a feature engineering layer, a prediction model layer, and a decision support layer, characterized by: The data collection layer is used to collect futures historical prices, trading volumes, open interest, news reports, investor discussion information, and macroeconomic indicator data from futures exchanges, news media, social platforms, government economic databases, and authoritative financial institution websites; The data preprocessing layer includes a data cleaning module, a data conversion module, and a data normalization module; The feature engineering layer includes a technical indicator feature extraction module, a sentiment feature extraction module, a macroeconomic feature integration module, and a composite feature construction module; The prediction model layer includes an LSTM model building module, a Transformer model integration module, a reinforcement learning optimization module, and a model fusion and tuning module; The decision support layer includes a transaction signal generation module, an explainability analysis and display module, and a risk assessment and prompt module.
2. The futures market forecasting system according to claim 1, characterized in that: In the data collection layer, futures exchange data is collected in real time through the API interface, news media and social platform data are collected regularly through web crawlers, and macroeconomic data is regularly obtained from relevant websites.
3. The futures market forecasting system according to claim 1, characterized in that: In the data preprocessing layer, the data cleaning module is used to identify and correct outliers and fill missing values. It uses the IQR algorithm to identify outliers and uses the time series interpolation method to fill missing values. The data conversion module is used to convert unstructured data into structured data; The data normalization module is used to eliminate the influence of different dimensions, and Z-score normalization or Min-Max normalization method is used for numerical data.
4. The futures market forecasting system according to claim 1, characterized in that: In the feature engineering layer, the technical indicator feature extraction module calculates the relative strength index, the moving average convergence divergence index, and the Bollinger band technical analysis index; The emotion feature extraction module uses the BERT pre-trained language model to perform sentiment analysis; The macroeconomic characteristic integration module associates and integrates macroeconomic indicator data with futures market data to construct a relationship model between macroeconomic factors and futures markets; The composite feature construction module constructs a composite feature vector and performs dimensionality reduction processing, and reduces the feature dimension through principal component analysis dimensionality reduction technology.
5. The futures market forecasting system according to claim 1, characterized in that: In the prediction model layer, the LSTM model building module is used to process the long-term dependency of time series; The Transformer model integration module is combined with the LSTM model to capture data associations at different time points; The reinforcement learning optimization module uses a proximal strategy to optimize the reinforcement learning algorithm and adjusts parameters based on real-time market feedback; The model fusion and tuning module adopts Stacking and Bagging model fusion technology, and uses cross-validation and grid search methods to tune the model parameters.
6. The futures market forecasting system according to claim 1, characterized in that: In the decision support layer, the transaction signal generation module generates transaction signals based on the prediction results combined with the preset transaction strategy, and the transaction signals include buy, sell or hold signals; The explainability analysis display module uses SHAP value analysis and LIME technology to display the model decision-making process and present the contribution of each feature to the prediction results to investors; The risk assessment and warning module uses the VaR risk value model to assess investment risks, and assesses investment risks and provides warnings based on forecast results and market fluctuations.