Stock quantitative analysis method and system combining big data and artificial intelligence

Abstract

The invention relates to the technical field of big data, and discloses a stock quantitative analysis method and system combining big data and artificial intelligence, and the method comprises the steps: determining a stock quantitative demand for quantifying a to-be-analyzed stock, carrying out the preprocessing of multi-source stock data, and obtaining the processed multi-source stock data; extracting and quantifying stock characteristics of the stock to be analyzed, calculating a sensitivity coefficient of the stock characteristics to the stock quantification demand, and screening target stock characteristics of the stock characteristics; performing multi-level dimensionality reduction on the target stock features to obtain dimensionality-reduced stock features, extracting advanced stock features of the dimensionality-reduced stock features, and analyzing and quantifying quantitative signals of the stocks to be analyzed; and analyzing the quantification risk of the quantification signal, performing parameter iteration on the quantification signal to obtain a target quantification signal, and performing stock quantification analysis for quantifying the to-be-analyzed stock based on the target quantification signal. According to the invention, the intellectualization of stock quantitative analysis can be improved.

Description

Technical Field

[0001] The present invention relates to a stock quantitative analysis method and system combining big data and artificial intelligence, and belongs to the technical field of big data. Background Art

[0002] Stock quantitative analysis refers to an investment method that uses mathematics, statistics, and computer technology to conduct stock analysis by constructing quantitative models and algorithms. Stock quantitative analysis is a systematic investment method based on data and models, with advantages such as objectivity, high efficiency, and risk controllability. With the development of big data and artificial intelligence technologies, quantitative analysis will play an increasingly important role in the future financial market.

[0003] Traditional ways of stock quantitative analysis usually rely on simple mathematical models and statistical methods to analyze historical price and trading volume data, and generate trading instructions through preset trading rules and signals. This way often only focuses on technical indicators and lacks in-depth analysis of stock data, resulting in an insufficiently intelligent stock quantitative analysis process. Summary of the Invention

[0004] The present invention provides a stock quantitative analysis method and system combining big data and artificial intelligence, and its main purpose is to improve the intelligence of stock quantitative analysis.

[0005] To achieve the above object, a stock quantitative analysis method combining big data and artificial intelligence provided by the present invention includes:

[0006] Determine the stock quantitative requirements of the stock to be quantitatively analyzed. According to the stock quantitative requirements, obtain multi-source stock data of the stock to be quantitatively analyzed, where the stock data includes historical market data, unstructured data, macro index data, and fundamental data, and preprocess the multi-source stock data to obtain processed multi-source stock data;

[0007] Extract the stock characteristics of the stock to be quantitatively analyzed according to the processed multi-source stock data, where the stock characteristics include technical indicators, fundamental indicators, market sentiment indicators, and macro indicators;

[0008] Calculate the sensitivity coefficient of the stock characteristics to the stock quantitative requirements, and screen the target stock characteristics of the stock characteristics according to the sensitivity coefficient;

[0009] According to the target stock features, use the dimensionality reduction network in the trained stock quantitative analysis model to perform multi-level dimensionality reduction on the target stock features to obtain dimensionality-reduced stock features. Use the feature extraction network in the stock quantitative analysis model to extract the high-level stock features of the dimensionality-reduced stock features. According to the high-level stock features, use the quantitative analysis network in the stock quantitative analysis model to analyze the quantitative signals of the stock to be quantitatively analyzed;

[0010] Use a preset intelligent order risk analysis algorithm to analyze the quantitative risk of the quantitative signal, and through the quantitative risk, perform parameter iteration on the quantitative signal to obtain a target quantitative signal. Based on the target quantitative signal, perform stock quantitative analysis on the stock to be quantitatively analyzed.

[0011] Optionally, the preprocessing of the multi-source stock data to obtain processed multi-source stock data includes:

[0012] Mark the missing values in the multi-source stock data, and supplement the missing values to obtain supplemented multi-source stock data;

[0013] Calculate the outliers of the supplemented multi-source stock data;

[0014] Correct the outliers to obtain corrected multi-source stock data;

[0015] Identify the stock codes and timestamps of the corrected multi-source stock data;

[0016] According to the stock codes and timestamps, integrate the corrected multi-source stock data into a preset data framework to obtain the processed multi-source stock data.

[0017] Optionally, the extraction of the stock features of the stock to be quantitatively analyzed according to the processed multi-source stock data includes:

[0018] Mark the technical indicator data of the processed multi-source stock data;

[0019] Through the technical indicator data, calculate the moving average, relative strength index, Bollinger Bands, and MACD of the stock to be quantitatively analyzed;

[0020] Based on the moving average, relative strength index, Bollinger Bands, and MACD, determine the technical indicators of the stock to be quantitatively analyzed;

[0021] Mark the fundamental data of the processed multi-source stock data, and through the fundamental data, calculate the price-to-earnings ratio and price-to-book ratio of the stock to be quantitatively analyzed;

[0022] Through the price-to-earnings ratio and price-to-book ratio, determine the fundamental indicators of the stock to be quantitatively analyzed;

[0023] Label the market sentiment data for processing multi-source stock data, and analyze the market sentiment indicators of the quantized stock to be analyzed according to the market sentiment data;

[0024] Label the macro-indicator data for processing multi-source stock data, and calculate the GDP growth rate and inflation rate of the quantized stock to be analyzed through the macro-indicator data;

[0025] Determine the macro-indicators of the quantized stock to be analyzed according to the GDP growth rate and inflation rate;

[0026] Combine the technical indicators, the fundamental indicators, the market sentiment indicators, and the macro-indicators to determine the stock characteristics of the quantized stock to be analyzed.

[0027] Optionally, the analyzing the market sentiment indicators of the quantized stock to be analyzed according to the market sentiment data includes:

[0028] Perform word segmentation on the market sentiment data to obtain segmented market sentiment data;

[0029] Remove the stop words in the segmented market sentiment data to obtain target market sentiment data;

[0030] Label the news and comments in the target market sentiment data;

[0031] Calculate the news sentiment indicator and the comment sentiment indicator of the news and comments;

[0032] Based on the news sentiment indicator and the comment sentiment indicator, use the following formula to calculate the market sentiment indicator of the quantized stock to be analyzed:

[0033] z = α·S x +(1 - α)·S p

[0034] where z represents the market sentiment indicator, S x represents the news sentiment indicator, S p represents the comment sentiment indicator, α represents the weight of the news sentiment indicator, and (1 - α) represents the weight of the comment sentiment indicator.

[0035] Optionally, the calculating the news sentiment indicator and the comment sentiment indicator of the news and comments includes:

[0036] Label the news sentiment words and comment sentiment words of the news and comments;

[0037] Analyze the news sentiment word intensity and comment sentiment word intensity of the news sentiment words and comment sentiment words;

[0038] Based on the news sentiment word intensity and the comment sentiment word intensity, use the following formula to calculate the news sentiment index and the comment sentiment index of the news and the comment:

[0039]

[0040] Among them, S x represents the news sentiment index, S p represents the comment sentiment index, m represents the number of news, ω c represents the news weight of the c-th news, ω c,i represents the news sentiment word intensity of the i-th news sentiment word in the c-th news, n represents the number of news sentiment words in the c-th news, r represents the number of comments, ω v represents the comment weight of the v-th comment, ω v,o represents the comment sentiment word intensity of the o-th comment sentiment word in the v-th comment, and u represents the number of comment sentiment words in the v-th comment.

[0041] Optionally, calculating the sensitivity coefficient of the stock feature to the stock quantification requirement includes:

[0042] Analyze the characteristic linear relationship of the stock feature;

[0043] Define the benchmark model of the stock feature according to the characteristic linear relationship;

[0044] Construct the interaction term of the stock feature in the benchmark model;

[0045] Use the benchmark model to calculate the stock feature partial derivative and the interaction term partial derivative of the stock feature and the interaction term;

[0046] Based on the stock feature partial derivative and the interaction term partial derivative, analyze the local effect and the global effect of the stock feature on the stock quantification requirement;

[0047] Calculate the sensitivity coefficient of the stock feature to the stock quantification requirement according to the local effect and the global effect.

[0048] Optionally, according to the target stock feature, using the dimensionality reduction network in the trained stock quantification analysis model to perform multi-level dimensionality reduction on the target stock feature to obtain the dimensionality reduction stock feature, including:

[0049] Identify the internal structure of the dimensionality reduction network;

[0050] Based on the internal structure, preprocess the target stock feature to obtain the processed target stock feature;

[0051] Define the dimensionality reduction levels of the dimensionality reduction network;

[0052] According to the dimensionality reduction level, use the hidden layer in the dimensionality reduction network to reduce the dimension of the processed target stock features to obtain initial dimensionality-reduced stock features;

[0053] Calculate the dimensionality reduction loss of the initial dimensionality-reduced stock features;

[0054] When the dimensionality reduction loss meets the preset dimensionality reduction loss threshold, use the initial dimensionality-reduced stock features as the dimensionality-reduced stock features.

[0055] Optionally, the analyzing the quantization signal of the stock to be analyzed by using the quantization analysis network in the stock quantization analysis model according to the high-level stock features includes:

[0056] Construct a feature matrix of the high-level stock features;

[0057] According to the feature matrix, use the convolutional layer in the quantization analysis network to extract the hidden features of the stock to be analyzed by quantization;

[0058] Based on the hidden features, calculate the quantization signal probability of the stock to be analyzed by quantization using the following formula:

[0059] y = softmax(W 2 H 2 + g)

[0060] where y represents the quantization signal probability of the stock to be analyzed by quantization, W 2 represents the weight matrix of the stock to be analyzed by quantization, H 2 represents the hidden features of the stock to be analyzed by quantization, g represents the bias term of the stock to be analyzed by quantization, and softmax represents the probability distribution function;

[0061] Determine the quantization signal of the stock to be analyzed by quantization through the quantization signal probability.

[0062] Optionally, the analyzing the quantization risk of the quantization signal by using a preset intelligent order risk analysis algorithm includes:

[0063] Define the risk tolerance of the quantization signal;

[0064] According to the risk tolerance, determine the algorithm parameters of the intelligent order risk analysis algorithm;

[0065] According to the algorithm parameters and the risk tolerance, use the intelligent order risk analysis algorithm to analyze the risk-return ratio of the quantization signal;

[0066] According to the risk-return ratio, determine the quantization risk of the quantization signal.

[0067] To solve the above problems, the present invention further provides a stock quantitative analysis system combining big data and artificial intelligence, and the system includes:

[0068] A stock data acquisition module, configured to determine the stock quantitative requirements for quantifying the stock to be analyzed, and according to the stock quantitative requirements, acquire multi-source stock data of the stock to be analyzed, wherein the stock data includes historical market data, unstructured data, macro index data, and fundamental data, and preprocess the multi-source stock data to obtain processed multi-source stock data;

[0069] A stock feature extraction module, configured to extract stock features of the stock to be analyzed according to the processed multi-source stock data, wherein the stock features include technical indicators, fundamental indicators, market sentiment indicators, and macro indicators;

[0070] A stock feature analysis module, configured to calculate the sensitivity coefficient of the stock features to the stock quantitative requirements, and according to the sensitivity coefficient, screen the target stock features of the stock features;

[0071] A quantitative signal generation module, configured to perform multi-level dimensionality reduction on the target stock features by using a dimensionality reduction network in the trained stock quantitative analysis model according to the target stock features to obtain dimensionality-reduced stock features, extract high-level stock features of the dimensionality-reduced stock features by using a feature extraction network in the stock quantitative analysis model, and analyze the quantitative signal of the stock to be analyzed by using a quantitative analysis network in the stock quantitative analysis model according to the high-level stock features;

[0072] A stock quantitative analysis module, configured to analyze the quantitative risk of the quantitative signal by using a preset intelligent order risk analysis algorithm, and perform parameter iteration on the quantitative signal through the quantitative risk to obtain a target quantitative signal, and perform stock quantitative analysis on the stock to be analyzed based on the target quantitative signal.

[0073] Compared with the problems described in the background art, first, through multi-source data fusion, the system comprehensively captures multi-dimensional information of the stock market, including historical market conditions, unstructured text, macroeconomic indicators, and fundamental data, ensuring the depth and breadth of analysis. This comprehensive data preprocessing provides a solid foundation for subsequent feature extraction and model training. Second, by calculating the sensitivity coefficients of stock features to quantitative requirements and screening out target stock features, the system realizes the efficient screening of key information, reduces noise interference, and improves the accuracy of analysis. The application of the multi-level dimensionality reduction network further simplifies the data structure and enhances the operation efficiency and prediction ability of the model. Third, using the trained feature extraction network and quantitative analysis network, the system can extract high-level stock features from complex data and accurately analyze quantitative signals, providing a scientific basis for investment decisions. This method reduces investment risks caused by subjective judgment errors. Finally, through the intelligent order risk analysis algorithm to analyze the risks of quantitative signals and combined with parameter iterative optimization, the system can generate target quantitative signals, thus realizing the feasibility under controllable risks. This iterative optimization process ensures the continuous adaptability and market competitiveness of the quantitative strategy. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 FIG. is a schematic flow chart of a stock quantitative analysis method combining big data and artificial intelligence provided by an embodiment of the present invention;

[0075] Figure 2 FIG. is a schematic block diagram of a module for implementing the stock quantitative analysis method combining big data and artificial intelligence provided by an embodiment of the present invention.

[0076] The implementation, functional features, and advantages of the present invention will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0077] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0078] An embodiment of the present application provides a stock quantitative analysis method combining big data and artificial intelligence. The execution subject of the stock quantitative analysis method combining big data and artificial intelligence includes, but is not limited to, at least one of electronic devices such as a server, a terminal, etc. that can be configured to execute the method provided by the embodiment of the present application. In other words, the stock quantitative analysis method combining big data and artificial intelligence can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to: a single server, a server cluster, a cloud server, or a cloud server cluster, etc.

[0079] Embodiment 1:

[0080] Refer to Figure 1As shown in the figure, it is a schematic flowchart of a stock quantitative analysis method combining big data and artificial intelligence provided by an embodiment of the present invention. In this embodiment, the stock quantitative analysis method combining big data and artificial intelligence includes:

[0081] S1. Determine the stock quantitative requirements of the stocks to be analyzed quantitatively. According to the stock quantitative requirements, obtain multi-source stock data of the stocks to be analyzed quantitatively. Among them, the stock data includes historical market data, unstructured data, macro index data, and fundamental data. Preprocess the multi-source stock data to obtain processed multi-source stock data.

[0082] It should be explained that the stocks to be analyzed quantitatively refer to specific stocks or stock portfolios selected as the objects of quantitative analysis. The stock quantitative requirements refer to the specific goals and requirements that need to be clarified and met during the quantitative analysis process, such as investment goals, risk preferences, and other requirements. The historical market data refers to the records of stocks over a period of time in the past. The unstructured data includes information such as news articles, social media, and financial reports. The macro index data refers to economic indicators that affect the entire market, such as GDP growth rate, unemployment rate, and other indicators. The fundamental data refers to data directly related to the company's financial status and business performance, such as net profit, price-earnings ratio, dividend rate, and other data.

[0083] The present invention preprocesses the multi-source stock data to obtain processed multi-source stock data, which can improve the quality and consistency of the data and lay a foundation for subsequent quantitative analysis and modeling.

[0084] Specifically, the preprocessing of the multi-source stock data to obtain processed multi-source stock data includes:

[0085] Mark the missing values of the multi-source stock data, and supplement the missing values to obtain supplemented multi-source stock data;

[0086] Calculate the outliers of the supplemented multi-source stock data;

[0087] Correct the outliers to obtain corrected multi-source stock data;

[0088] Identify the stock codes and timestamps of the corrected multi-source stock data;

[0089] According to the stock codes and timestamps, integrate the corrected multi-source stock data into a preset data framework to obtain the processed multi-source stock data.

[0090] Among them, the missing value refers to certain fields with empty data values. The supplemented multi-source stock data refers to the data set after filling in the missing values. The outlier refers to the value in the data set that significantly deviates from the normal range. The corrected multi-source stock data refers to the data set after correcting the outliers. The stock code refers to the code that uniquely identifies a stock. The timestamp refers to the time point when the data is recorded. The processed multi-source stock data refers to the final data set after supplementing missing values, correcting outliers, and data integration.

[0091] Optionally, filling in the missing values to obtain the supplemented multi-source stock data can be achieved by interpolation methods (such as linear interpolation, mean interpolation).

[0092] Optionally, detecting the outliers of the supplemented multi-source stock data can be performed through the 3σ principle.

[0093] S2. According to the processed multi-source stock data, extract the stock features of the stock to be analyzed quantitatively. Among them, the stock features include technical indicators, fundamental indicators, market sentiment indicators, and macro indicators.

[0094] The present invention can construct a comprehensive stock feature set by extracting the stock features of the stock to be analyzed quantitatively according to the processed multi-source stock data.

[0095] Specifically, extracting the stock features of the stock to be analyzed quantitatively according to the processed multi-source stock data includes:

[0096] Mark the technical indicator data of the processed multi-source stock data;

[0097] Based on the technical indicator data, calculate the moving average, relative strength index, Bollinger Bands, and MACD of the stock to be analyzed quantitatively;

[0098] Based on the moving average, relative strength index, Bollinger Bands, and MACD, determine the technical indicators of the stock to be analyzed quantitatively;

[0099] Mark the fundamental data of the processed multi-source stock data, and based on the fundamental data, calculate the price-earnings ratio and price-to-book ratio of the stock to be analyzed quantitatively;

[0100] Based on the price-earnings ratio and price-to-book ratio, determine the fundamental indicators of the stock to be analyzed quantitatively;

[0101] Mark the market sentiment data of the processed multi-source stock data, and analyze the market sentiment indicators of the stock to be analyzed quantitatively according to the market sentiment data;

[0102] Mark the macro - indicator data for processing multi - source stock data, and calculate the GDP growth rate and inflation rate of the stock to be analyzed quantitatively based on the macro - indicator data;

[0103] Determine the macro - indicators of the stock to be analyzed quantitatively according to the GDP growth rate and inflation rate;

[0104] Combine the technical indicators, the fundamental indicators, the market sentiment indicators and the macro - indicators to determine the stock characteristics of the stock to be analyzed quantitatively.

[0105] Among them, the technical - indicator data refers to historical data such as stock price and trading volume. The moving average refers to the average value of a stock over a period of time. The relative strength index refers to an indicator that measures the intensity of stock - price changes. The Bollinger Bands refer to parameters that measure the price - volatility range. The MACD refers to the difference between short - term and long - term exponentially - weighted moving averages. The technical indicator is a comprehensive indicator calculated through moving averages, RSI, Bollinger Bands and MACD, etc. The fundamental data refers to data related to the company's financial and operating conditions, such as earnings per share, net assets per share, etc. The price - earnings ratio is the ratio of the stock price to earnings per share, which is used to measure the valuation level of a stock. The price - to - book ratio is the ratio of the stock price to net assets per share, which is used to measure the value of a stock. The fundamental indicator is a comprehensive indicator calculated through the price - earnings ratio and price - to - book ratio, etc. The market - sentiment data refers to text data that reflects the emotions of market participants, such as news, social - media comments, etc. The market - sentiment indicator is a comprehensive indicator obtained by performing sentiment analysis on news and social - media data. The macro - indicator data refers to data that reflects the overall economic environment. The GDP growth rate is an indicator that measures the economic growth rate. The inflation rate is an indicator that measures the change in price levels. The macro - indicator is a comprehensive indicator calculated through the GDP growth rate and inflation rate, etc. The stock characteristics refer to a comprehensive characteristic set that combines technical indicators, fundamental indicators, market - sentiment indicators and macro - indicators.

[0106] Further, the analysis of the market - sentiment indicator of the stock to be analyzed quantitatively based on the market - sentiment data includes:

[0107] Perform word - segmentation on the market - sentiment data to obtain word - segmented market - sentiment data;

[0108] Remove stop - words from the word - segmented market - sentiment data to obtain target market - sentiment data;

[0109] Mark the news and comments in the target market - sentiment data;

[0110] Calculate the news - sentiment indicator and comment - sentiment indicator of the news and comments;

[0111] Based on the news sentiment index and the comment sentiment index, use the following formula to calculate the market sentiment index of the stock to be analyzed:

[0112] z = α·S x +(1 - α)·S p

[0113] Where z represents the market sentiment index, S x represents the news sentiment index, S p represents the comment sentiment index, α represents the weight of the news sentiment index, and (1 - α) represents the weight of the comment sentiment index.

[0114] Among them, the segmented market sentiment data refers to the data obtained by segmenting the original market sentiment text (such as news reports, social media comments, etc.). The stop words refer to those words that appear frequently but contribute little to the meaning of the text, such as "de", "he", "shi", etc. The target market sentiment data refers to the set of valuable vocabulary remaining after removing the stop words. The news refers to the news report text related to the stock to be analyzed, and the comment refers to the social media comments or other user-generated comment texts related to the stock to be analyzed. The news sentiment index refers to the sentiment score calculated based on the news content, reflecting the sentiment tendency of the news report towards the stock to be analyzed. The comment sentiment index refers to the sentiment score calculated based on the comment content, reflecting the sentiment tendency of the user comment towards the stock to be analyzed.

[0115] Furthermore, calculating the news sentiment index and the comment sentiment index of the news and comments includes:

[0116] Mark the news sentiment words and comment sentiment words in the news and comments;

[0117] Analyze the news sentiment word intensity and comment sentiment word intensity of the news sentiment words and comment sentiment words;

[0118] Based on the news sentiment word intensity and comment sentiment word intensity, use the following formula to calculate the news sentiment index and the comment sentiment index of the news and comments:

[0119]

[0120]

[0121] Where S x represents the news sentiment index, S p represents the comment sentiment index, m represents the number of news, ω c represents the news weight of the c-th news, ω c,iDenote the intensity of the news sentiment word of the \(i\)-th news sentiment word in the \(c\)-th news, \(n\) represents the number of news sentiment words in the \(c\)-th news, \(r\) represents the number of comments, and \(\omega\) v Denote the comment weight of the \(v\)-th comment, \(\omega\) v,o Denote the intensity of the comment sentiment word of the \(o\)-th comment sentiment word in the \(v\)-th comment, and \(u\) represents the number of comment sentiment words in the \(v\)-th comment.

[0122] Among them, the news sentiment word refers to the word expressing emotion in the news text (such as "rise", "plunge", "optimistic", "pessimistic"), the comment sentiment word refers to the word expressing emotion in the comment text (such as "favorable", "junk", "bull market", "bear market"), the news sentiment word intensity refers to the weight value of the news sentiment word in the sentiment dictionary, and the comment sentiment word intensity refers to the weight value of the comment sentiment word in the sentiment dictionary.

[0123] S3. Calculate the sensitivity coefficient of the stock feature to the stock quantification demand, and according to the sensitivity coefficient, screen the target stock feature of the stock feature.

[0124] The present invention can provide a basis for later feature selection by calculating the sensitivity coefficient of the stock feature to the stock quantification demand.

[0125] Specifically, the calculation of the sensitivity coefficient of the stock feature to the stock quantification demand includes:

[0126] Analyze the feature linear relationship of the stock feature;

[0127] According to the feature linear relationship, define the benchmark model of the stock feature;

[0128] Construct the interaction term of the stock feature in the benchmark model;

[0129] Using the benchmark model, calculate the stock feature partial derivative and the interaction term partial derivative of the stock feature and the interaction term;

[0130] Based on the stock feature partial derivative and the interaction term partial derivative, analyze the local effect and the global effect of the stock feature on the stock quantification demand;

[0131] According to the local effect and the global effect, calculate the sensitivity coefficient of the stock feature to the stock quantification demand.

[0132] Among them, the described characteristic linear relationship refers to the linear correlation between stock characteristics and between stock characteristics and quantitative requirements. The benchmark model refers to a model used to evaluate the impact of stock characteristics on quantitative requirements. The interaction term refers to a special term in the model, indicating the impact of the interaction between two or more characteristics on quantitative requirements. The partial derivative of the stock characteristic refers to the partial derivative of the model output (quantitative requirement) with respect to a specific stock characteristic. The partial derivative of the interaction term refers to the partial derivative of the model output with respect to the interaction term. The local effect refers to the direct impact of the change in a single stock characteristic or interaction term on quantitative requirements. The global effect refers to the overall impact of the combined changes of all stock characteristics and interaction terms on quantitative requirements. The sensitivity coefficient refers to a quantitative indicator for measuring the impact of changes in stock characteristics on quantitative requirements.

[0133] Optionally, analyzing the local effect and global effect of the stock characteristics on the stock quantitative requirements based on the partial derivative of the stock characteristic and the partial derivative of the interaction term can be analyzed through a local influence diagram and a global interpretation model.

[0134] It should be explained that the target stock characteristic refers to the stock characteristic that is considered to have a significant impact on stock quantitative requirements after analysis.

[0135] S4. According to the target stock characteristics, use the dimensionality reduction network in the trained stock quantitative analysis model to perform multi-level dimensionality reduction on the target stock characteristics to obtain dimensionality-reduced stock characteristics. Use the feature extraction network in the stock quantitative analysis model to extract the high-level stock characteristics of the dimensionality-reduced stock characteristics. According to the high-level stock characteristics, use the quantitative analysis network in the stock quantitative analysis model to analyze the quantitative signals of the stocks to be analyzed quantitatively.

[0136] According to the target stock characteristics of the present invention, use the dimensionality reduction network in the trained stock quantitative analysis model to perform multi-level dimensionality reduction on the target stock characteristics to obtain dimensionality-reduced stock characteristics to ensure that the most important characteristics for stock quantitative analysis are retained.

[0137] Specifically, according to the target stock characteristics, using the dimensionality reduction network in the trained stock quantitative analysis model to perform multi-level dimensionality reduction on the target stock characteristics to obtain dimensionality-reduced stock characteristics includes:

[0138] Identify the internal structure of the dimensionality reduction network;

[0139] Based on the internal structure, preprocess the target stock characteristics to obtain processed target stock characteristics;

[0140] Define the dimensionality reduction levels of the dimensionality reduction network;

[0141] According to the dimensionality reduction level, use the hidden layer in the dimensionality reduction network to reduce the dimension of the processed target stock features to obtain initial reduced-dimension stock features;

[0142] Calculate the dimensionality reduction loss of the initial reduced-dimension stock features;

[0143] When the dimensionality reduction loss meets the preset dimensionality reduction loss threshold, use the initial reduced-dimension stock features as the reduced-dimension stock features.

[0144] Among them, the internal structure refers to the architectural details of the dimensionality reduction network, including the number of layers, the number of neurons in each layer, the activation function, the connection method (such as full connection, convolution, etc.), and any specific network design elements (such as autoencoder, bottleneck layer, etc.). The processed target stock features refer to the target stock features after necessary data preprocessing (such as standardization, normalization, missing value processing, etc.). The dimensionality reduction level refers to each level in the network for implementing feature dimensionality reduction. The hidden layer refers to the intermediate layer in the network. The initial reduced-dimension stock features refer to the feature set obtained by the first dimensionality reduction after the data passes through the hidden layer of the dimensionality reduction network. The dimensionality reduction loss refers to the quantitative index of information loss during the dimensionality reduction process. The dimensionality reduction loss threshold refers to the parameter used to determine whether the dimensionality reduction loss is within the acceptable range. The reduced-dimension stock features refer to the feature set that finally passes through the dimensionality reduction network processing and reaches the preset dimensionality reduction loss threshold or meets other termination conditions.

[0145] The present invention uses the feature extraction network in the stock quantitative analysis model to extract the high-level stock features of the reduced-dimension stock features. According to the high-level stock features, more valuable high-level features can be extracted from the reduced-dimension stock features by using the feature extraction network, thereby improving the performance of the stock quantitative analysis model and the quality of investment decisions. Among them, the high-level stock features refer to the features with stronger interpretability extracted from the reduced-dimension stock features through the complex feature extraction network in stock quantitative analysis.

[0146] The present invention analyzes the quantization signal of the stock to be analyzed by using the quantization analysis network in the stock quantitative analysis model according to the high-level stock features, which can effectively use the quantization analysis network to deeply analyze the quantization signal of the stock to be analyzed and provide support for investment decisions.

[0147] Specifically, the analyzing the quantization signal of the stock to be analyzed by using the quantization analysis network in the stock quantitative analysis model according to the high-level stock features includes:

[0148] Construct the feature matrix of the high-level stock features;

[0149] According to the feature matrix, use the convolutional layer in the quantization analysis network to extract the hidden features of the stock to be analyzed;

[0150] Based on the hidden feature, calculate the quantization signal probability of the stock to be analyzed using the following formula:

[0151] y = softmax(W 2 H 2 + g)

[0152] where y represents the quantization signal probability of the stock to be analyzed, W 2 represents the weight matrix of the stock to be analyzed, H 2 represents the hidden feature of the stock to be analyzed, g represents the bias term of the stock to be analyzed, and softmax represents the probability distribution function;

[0153] Determine the quantization signal of the stock to be analyzed through the quantization signal probability.

[0154] Among them, the feature matrix refers to the matrix formed by arranging the feature values of each sample row by row. The convolutional layer is a layer structure in the quantization analysis network, which is used to extract local features from the input data. The hidden feature refers to the feature extracted from the original feature matrix through the convolutional layer. The quantization signal probability refers to the probability that the stock to be analyzed belongs to a certain category. The weight matrix refers to the neuron parameter that determines how the hidden feature is converted into the quantization signal probability. The bias term refers to the parameter used to adjust the output of the activation function. The probability distribution function refers to the function that converts the linear output of the neural network into probability values. The quantization signal refers to the final decision made based on the quantization signal probability.

[0155] S5. Analyze the quantization risk of the quantization signal using a preset intelligent order risk analysis algorithm, and perform parameter iteration on the quantization signal through the quantization risk to obtain a target quantization signal, and perform stock quantization analysis on the stock to be analyzed based on the target quantization signal.

[0156] The present invention analyzes the quantization risk of the quantization signal using a preset intelligent order risk analysis algorithm, which can help investors better manage risks when performing quantitative trading and ensure that trading decisions are based on a comprehensive risk assessment.

[0157] Specifically, the analysis of the quantization risk of the quantization signal using a preset intelligent order risk analysis algorithm includes:

[0158] Define the risk tolerance of the quantization signal;

[0159] Determine the algorithm parameters of the intelligent order risk analysis algorithm according to the risk tolerance;

[0160] Analyze the risk - return ratio of the quantitative signal using the intelligent order risk analysis algorithm according to the algorithm parameters and the risk tolerance;

[0161] Determine the quantitative risk of the quantitative signal according to the risk - return ratio.

[0162] Among them, the risk tolerance refers to the maximum potential loss level that an investor is willing to accept during the investment process. The algorithm parameters refer to a series of variables used to configure and optimize the intelligent order risk analysis algorithm, such as stop - loss, take - profit levels, risk management rules and other parameters. The risk - return ratio is an index that measures the relationship between potential returns and potential risks. The quantitative risk refers to the probability of losses that a stock may face.

[0163] Through the quantitative risk, the present invention performs parameter iteration on the quantitative signal to obtain a target quantitative signal, which can continuously optimize the parameters of the quantitative signal to obtain a better quantitative target. Among them, the parameter iteration of the quantitative signal can optimize the parameters through the crossover, mutation and selection processes of the genetic algorithm. The parameter iteration of the quantitative signal mainly iterates on parameters such as stop - loss points, take - profit points, entry conditions, and exit conditions.

[0164] Compared with the problems in the background technology, first, through multi - source data fusion, the system comprehensively captures multi - dimensional information of the stock market, including historical market conditions, unstructured text, macro - economic indicators and fundamental data, ensuring the depth and breadth of the analysis. This comprehensive data pre - processing provides a solid foundation for subsequent feature extraction and model training. Second, by calculating the sensitivity coefficient of stock features to quantitative requirements and screening out target stock features, the system realizes the efficient screening of key information, reduces noise interference, and improves the accuracy of the analysis. The application of the multi - level dimensionality reduction network further simplifies the data structure, improves the operation efficiency and prediction ability of the model. Third, using the trained feature extraction network and quantitative analysis network, the system can extract high - level stock features from complex data and accurately analyze quantitative signals, providing a scientific basis for investment decisions. This method reduces the investment risk caused by subjective judgment errors. Finally, through the risk analysis of quantitative signals using the intelligent order risk analysis algorithm and combined with parameter iteration optimization, the system can generate target quantitative signals, thus realizing the feasibility under controllable risks. This iterative optimization process ensures the continuous adaptability and market competitiveness of the quantitative strategy.

[0165] Embodiment 2:

[0166] As Figure 2 shown, it is a functional module diagram of a stock quantitative analysis system combining big data and artificial intelligence according to the present invention.

[0167] The stock quantitative analysis system 200 combining big data and artificial intelligence according to the present invention can be installed in an electronic device. According to the functions achieved, the stock quantitative analysis system combining big data and artificial intelligence may include a stock data acquisition module 201, a stock feature extraction module 202, a stock feature analysis module 203, a quantitative signal generation module 204, and a stock quantitative analysis module 205. The modules in the present invention may also be referred to as units, which refer to a series of computer program segments that can be executed by the processor of the electronic device and can complete fixed functions, and are stored in the memory of the electronic device.

[0168] In the embodiments of the present invention, the functions of each module / unit are as follows:

[0169] The stock data acquisition module 201 is used to determine the stock quantitative requirements for the stock to be analyzed quantitatively, and according to the stock quantitative requirements, obtain multi-source stock data of the stock to be analyzed quantitatively, where the stock data includes historical market data, unstructured data, macro index data, and fundamental data, and preprocess the multi-source stock data to obtain processed multi-source stock data;

[0170] The stock feature extraction module 202 is used to extract the stock features of the stock to be analyzed quantitatively according to the processed multi-source stock data, where the stock features include technical indicators, fundamental indicators, market sentiment indicators, and macro indicators;

[0171] The stock feature analysis module 203 is used to calculate the sensitivity coefficient of the stock features to the stock quantitative requirements, and according to the sensitivity coefficient, screen the target stock features of the stock features;

[0172] The quantitative signal generation module 204 is used to perform multi-level dimensionality reduction on the target stock features by using the dimensionality reduction network in the trained stock quantitative analysis model according to the target stock features to obtain dimensionality-reduced stock features, extract the high-level stock features of the dimensionality-reduced stock features by using the feature extraction network in the stock quantitative analysis model, and analyze the quantitative signals of the stock to be analyzed quantitatively by using the quantitative analysis network in the stock quantitative analysis model according to the high-level stock features;

[0173] The stock quantitative analysis module 205 is used to analyze the quantitative risk of the quantitative signal by using a preset intelligent order risk analysis algorithm, and perform parameter iteration on the quantitative signal through the quantitative risk to obtain a target quantitative signal, and perform stock quantitative analysis on the stock to be analyzed quantitatively based on the target quantitative signal.

[0174] Specifically, each module in the stock quantitative analysis system 200 combining big data and artificial intelligence in the embodiments of the present invention adopts the same as the above-mentionedFigure 1 The technical means are the same as those of the stock quantitative analysis method combining big data and artificial intelligence described in [reference], and can produce the same technical effects, which will not be elaborated here.

[0175] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms.

[0176] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A stock quantitative analysis method combining big data and artificial intelligence, characterized in that, The method includes: Determine the stock quantification requirements for the stock to be analyzed, and according to the stock quantification requirements, obtain multi-source stock data of the stock to be analyzed. Among them, the stock data includes historical market data, unstructured data, macro-indicator data, and fundamental data. Preprocess the multi-source stock data to obtain processed multi-source stock data; Extract the stock features of the stock to be analyzed according to the processed multi-source stock data. Among them, the stock features include technical indicators, fundamental indicators, market sentiment indicators, and macro-indicators; Calculate the sensitivity coefficient of the stock features to the stock quantification requirements, and according to the sensitivity coefficient, screen the target stock features of the stock features; According to the target stock features, use the dimensionality reduction network in the trained stock quantification analysis model to perform multi-level dimensionality reduction on the target stock features to obtain dimensionality-reduced stock features. Use the feature extraction network in the stock quantification analysis model to extract the high-level stock features of the dimensionality-reduced stock features. According to the high-level stock features, use the quantification analysis network in the stock quantification analysis model to analyze the quantification signal of the stock to be analyzed; Analyze the quantification risk of the quantification signal using a preset intelligent order risk analysis algorithm, and through the quantification risk, perform parameter iteration on the quantification signal to obtain a target quantification signal, and perform stock quantification analysis of the stock to be analyzed based on the target quantification signal.

2. The stock quantitative analysis method combining big data and artificial intelligence according to claim 1, wherein The preprocessing of the multi-source stock data to obtain processed multi-source stock data includes: Mark the missing values of the multi-source stock data, and supplement the missing values to obtain supplemented multi-source stock data; Calculate the outliers of the supplemented multi-source stock data; Correct the outliers to obtain corrected multi-source stock data; Identify the stock code and timestamp of the corrected multi-source stock data; According to the stock code and timestamp, integrate the corrected multi-source stock data into a preset data framework to obtain the processed multi-source stock data.

3. The stock quantitative analysis method combining big data and artificial intelligence according to claim 2, wherein The extraction of the stock features of the stock to be analyzed according to the processed multi-source stock data includes: Mark the technical indicator data of the processed multi-source stock data; Through the technical indicator data, calculate the moving average, relative strength index, Bollinger Bands, and MACD of the stock to be analyzed; Based on the moving average, relative strength index, Bollinger Bands, and MACD, determine the technical indicators of the stock to be analyzed; Mark the fundamental data of the processed multi-source stock data, and through the fundamental data, calculate the price-earnings ratio and price-to-book ratio of the stock to be analyzed; Through the price-earnings ratio and price-to-book ratio, determine the fundamental indicators of the stock to be analyzed; Mark the market sentiment data of the processed multi-source stock data, and analyze the market sentiment indicators of the stock to be analyzed according to the market sentiment data; Mark the macro-indicator data of the processed multi-source stock data, and through the macro-indicator data, calculate the GDP growth rate and inflation rate of the stock to be analyzed; Determine the macro indicators for quantifying the stock to be analyzed according to the GDP growth rate and inflation rate; Combine the technical indicators, fundamental indicators, market sentiment indicators, and macro indicators to determine the stock characteristics of the stock to be analyzed.

4. The stock quantitative analysis method combining big data and artificial intelligence according to claim 3, characterized in that, The analysis of the market sentiment indicators of the stock to be analyzed according to the market sentiment data includes: Perform word segmentation on the market sentiment data to obtain segmented market sentiment data; Remove stop words from the segmented market sentiment data to obtain target market sentiment data; Mark the news and comments in the target market sentiment data; Calculate the news sentiment indicators and comment sentiment indicators of the news and comments; Based on the news sentiment indicators and comment sentiment indicators, use the following formula to calculate the market sentiment indicators of the stock to be analyzed: z = α·S x +(1 - α)·S p Among them, z represents the market sentiment index, and S x represents the news sentiment index, and S p represents the comment sentiment index. α represents the weight of the news sentiment index, and (1-α) represents the weight of the comment sentiment index.

5. The stock quantitative analysis method combining big data and artificial intelligence according to claim 4, characterized in that The calculation of the news sentiment indicators and comment sentiment indicators of the news and comments includes: Mark the news sentiment words and comment sentiment words in the news and comments; Analyze the news sentiment word intensity and comment sentiment word intensity of the news sentiment words and comment sentiment words; Based on the news sentiment word intensity and comment sentiment word intensity, use the following formula to calculate the news sentiment indicators and comment sentiment indicators of the news and comments: Among them, S x represents the news sentiment index, S p represents the comment sentiment index, m represents the number of news, ω c represents the news weight of the c-th news, ω c,i represents the news sentiment word intensity of the i-th news sentiment word in the c-th news, n represents the number of news sentiment words in the c-th news, r represents the number of comments, ω v represents the comment weight of the v-th comment, ω v,o represents the comment sentiment word intensity of the o-th comment sentiment word in the v-th comment, u represents the number of comment sentiment words in the v-th comment.

6. The stock quantitative analysis method combining big data and artificial intelligence according to claim 5, characterized in that, The calculation of the sensitivity coefficient of the stock characteristics to the stock quantification requirements includes: Analyze the characteristic linear relationship of the stock characteristics; Define the benchmark model of the stock characteristics according to the characteristic linear relationship; Construct the interaction term of the stock characteristics in the benchmark model; Use the benchmark model to calculate the stock characteristic partial derivatives and interaction term partial derivatives of the stock characteristics and the interaction term; Based on the stock characteristic partial derivatives and interaction term partial derivatives, analyze the local effect and global effect of the stock characteristics on the stock quantification requirements; Calculate the sensitivity coefficient of the stock characteristics to the stock quantification requirements according to the local effect and global effect.

7. The stock quantitative analysis method combining big data and artificial intelligence according to claim 6, characterized in that The multi-level dimensionality reduction of the target stock characteristics by using the dimensionality reduction network in the trained stock quantification analysis model according to the target stock characteristics to obtain the dimensionality-reduced stock characteristics includes: Identify the internal structure of the dimensionality reduction network; Based on the internal structure, preprocess the target stock characteristics to obtain the processed target stock characteristics; Define the dimensionality reduction levels of the dimensionality reduction network; According to the dimensionality reduction levels, use the hidden layer in the dimensionality reduction network to perform dimensionality reduction on the processed target stock characteristics to obtain the initial dimensionality-reduced stock characteristics; Calculate the dimensionality reduction loss of the initial dimensionality-reduced stock characteristics; When the dimensionality reduction loss meets the preset dimensionality reduction loss threshold, use the initial dimensionality-reduced stock characteristics as the dimensionality-reduced stock characteristics.

8. The stock quantitative analysis method combining big data and artificial intelligence according to claim 7, wherein, The analysis of the quantization signal of the stock to be analyzed by using the quantization analysis network in the stock quantification analysis model according to the high-level stock characteristics includes: Construct the feature matrix of the high-level stock characteristics; According to the feature matrix, use the convolutional layer in the quantization analysis network to extract the hidden features of the stock to be analyzed; Based on the hidden features, use the following formula to calculate the quantization signal probability of the stock to be analyzed: y = softmax(W 2 H 2 + g) Among them, y represents the quantization signal probability of the stock to be analyzed, W 2 represents the weight matrix of the stock to be analyzed, H 2 represents the hidden features of the stock to be analyzed, g represents the bias term of the stock to be analyzed, and softmax represents the probability distribution function; Determine the quantization signal of the stock to be analyzed by means of the quantization signal probability.

9. The stock quantitative analysis method combining big data and artificial intelligence according to claim 8, wherein, Analyzing the quantization risk of the quantization signal by using a preset intelligent order risk analysis algorithm includes: Defining the risk tolerance of the quantization signal; Determining the algorithm parameters of the intelligent order risk analysis algorithm according to the risk tolerance; Analyzing the risk-return ratio of the quantization signal by using the intelligent order risk analysis algorithm according to the algorithm parameters and the risk tolerance; Determining the quantization risk of the quantization signal according to the risk-return ratio.

10. A stock quantitative analysis system combining big data and artificial intelligence, characterized in that, The system includes: A stock data acquisition module, configured to determine the stock quantization requirements of the stock to be analyzed, and acquire multi-source stock data of the stock to be analyzed according to the stock quantization requirements, where the stock data includes historical market data, unstructured data, macro indicator data, and fundamental data, and preprocess the multi-source stock data to obtain processed multi-source stock data; A stock feature extraction module, configured to extract the stock features of the stock to be analyzed according to the processed multi-source stock data, where the stock features include technical indicators, fundamental indicators, market sentiment indicators, and macro indicators; A stock feature analysis module, configured to calculate the sensitivity coefficient of the stock features to the stock quantization requirements, and screen the target stock features of the stock features according to the sensitivity coefficient; A quantization signal generation module, configured to perform multi-level dimensionality reduction on the target stock features by using a dimensionality reduction network in a trained stock quantization analysis model according to the target stock features to obtain dimensionality-reduced stock features, extract high-level stock features of the dimensionality-reduced stock features by using a feature extraction network in the stock quantization analysis model, and analyze the quantization signal of the stock to be analyzed by using a quantization analysis network in the stock quantization analysis model according to the high-level stock features; A stock quantization analysis module, configured to analyze the quantization risk of the quantization signal by using a preset intelligent order risk analysis algorithm, perform parameter iteration on the quantization signal through the quantization risk to obtain a target quantization signal, and perform stock quantization analysis of the stock to be analyzed based on the target quantization signal.

Images