CN105956770A - Stock market risk prediction platform and text excavation method thereof - Google Patents

Abstract

The invention discloses a stock market risk prediction platform, comprising: a data collection module; a data preprocessing module; a text mining module; a stock market prediction module; a risk assessment module; and a result output module. The present invention also provides a text mining method for a stock market risk prediction platform, which is a method for converting unstructured text data into structured data to analyze the viewpoints, attitudes or sentiments contained in the documents. The invention has a reasonable design, and converts unstructured text data into structured data to analyze the views, attitudes or emotions contained in the document, and evaluates the risk level of the stock market according to the results obtained by data analysis. The risk level of the stock market is not only It can serve investors' decision-making, and can also provide a basis for the government to formulate relevant policies and enterprises to implement corresponding strategies.

Description

A stock market risk prediction platform and its text mining method

technical field

The invention belongs to the field of stock market prediction and risk identification, and in particular relates to a stock market risk prediction platform and a text mining method thereof.

Background technique

The stock market is a barometer of a country or region's economic and financial activities, and it is also an important means of corporate financing and investor asset allocation. The forecast research on the stock market can not only provide a basis for the government, enterprises and investors to make relevant decisions, but also avoid Financial risks, and promote the stable and healthy development of the stock market.

The existing stock market prediction methods include securities investment analysis method, mathematical statistical model, nonlinear dynamics method, neural network, support vector machine, etc. These methods all assume that investors are rational and can conduct trading activities according to the principle of maximum utility. Nowadays, stock market activities are more complex and changeable. With the continuous discovery of financial anomalies such as herd effect, over-reaction or under-reaction, the defects of traditional forecasting methods are gradually highlighted.

In addition, with the development of information technology, the Internet contains massive amounts of information, including not only stock market transactions and other news, but also macroeconomic news, government-related policies and other content that have an important impact on the stock market. It has become a must for investors to obtain information. alternative channels. On the other hand, with the emergence of self-media and communication platforms such as forums and Weibo, stockholders express their opinions and exchange information on market trends, macroeconomic policies, investment intentions, etc. on the Internet, and the Internet has become a platform for mining investor sentiment. important carrier.

Most of the existing stock market forecasting platforms are based on traditional stock market forecasting methods, and their shortcomings are mainly reflected in the following three aspects:

First, the impact of investor sentiment and behavior on the stock market is ignored, and the forecast results cannot reflect the real market dynamics.

Second, focus on researching information such as stock market transactions, while ignoring the research on Internet news, forums and other data.

Third, there is a lack of risk assessment modules. The purpose of stock market forecasting is not only to guide investors in making decisions and obtain investment returns, but also to identify financial market risks, prevent systemic risks, and maintain financial market stability and national financial market security.

Contents of the invention

The technical problem to be solved by the present invention is to overcome the above-mentioned defects, provide a stock market risk prediction platform and its text mining method, with reasonable design, and a method for converting unstructured text data into structured data to analyze the viewpoints contained in the document , attitudes or emotions, and assess the risk level of the stock market based on the results of data analysis. The risk level of the stock market can not only serve investors' decision-making, but also provide a basis for the government to formulate relevant policies and enterprises to implement corresponding strategies.

In order to solve the above problems, the technical solution adopted in the present invention is:

A stock market risk prediction platform, characterized in that: comprising:

The data collection module is used to automatically collect and obtain stock market transaction data and multi-source Internet text data;

The data preprocessing module preprocesses the data obtained in the data acquisition module, including data cleaning, data integration, data transformation and data reduction, and prepares data for the establishment of a stock market prediction model;

The text mining module is used to analyze and process Internet text data to mine investor sentiment and construct a sentiment index, including six steps: text segmentation, part-of-speech tagging, emotional polarity tagging, sentiment index calculation, sentiment index adjustment, and sentiment index integration;

The stock market prediction module comprehensively applies the methods of text mining, machine learning and mathematical statistics to predict and analyze the stock market;

The risk assessment module divides the risk levels of the real-time monitored stocks and the overall market trend according to the results of the stock market forecast module;

The result output module is used to output the risk level of the stocks concerned to investors, and at the same time output the risk level of the entire market and provide real-time early warning.

The present invention also provides a text mining method for a stock market risk prediction platform, which is a method for converting unstructured text data into structured data to analyze the views, attitudes or emotions contained in the document;

The Internet text database used in the text mining method includes three aspects: policy news, financial news, and forum data. Policy news can mine the attitude and tendency of the government, financial news can understand the comprehensive information of social economy, and forum data can directly extract investor information. mood;

The text mining module in the stock market risk forecasting platform is to apply the text mining method to analyze and process the text data in the Internet, so as to extract the views, attitudes, and emotions of investors, and then use the calculated emotional index as an input variable to apply in stock market forecasting. module.

Due to the adoption of the above technical solution, compared with the prior art, the present invention has a reasonable design, a method for converting unstructured text data into structured data to analyze the views, attitudes or emotions contained in the document, and according to the data analysis The obtained results are used to assess the risk level of the stock market. The risk level of the stock market can not only serve investors' decision-making, but also provide a basis for the government to formulate relevant policies and enterprises to implement corresponding strategies.

At the same time, the present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

Description of drawings

Fig. 1 is the structural block diagram of stock market risk prediction platform in a kind of embodiment of the present invention;

Fig. 2 is the structural block diagram of stock market risk prediction platform module in a kind of embodiment of the present invention;

Fig. 3 is a flowchart of a text mining method in an embodiment of the present invention.

detailed description

Example:

A stock market risk prediction platform, as shown in Figure 1 and Figure 2, includes:

The data acquisition module, the built-in crawler program of the application platform automatically acquires the text data and stock market transaction data of the China Securities Regulatory Commission, the China Banking Regulatory Commission, the Central Bank, News Network, Hexun, Oriental Fortune, Sina Finance Forum, Netease Finance Forum, and Tencent Finance Forum.

The data preprocessing module performs denoising operations on the collected text data, including data cleaning, data integration, data transformation and data reduction, etc., to meet the needs of modeling.

The text mining module obtains the policy sentiment daily index, the financial sentiment daily index, the forum sentiment daily index and the comprehensive sentiment daily index according to the above text mining steps.

The stock market prediction module uses the comprehensive sentiment daily index and its lagged items, the Shanghai Composite Index rate of return and its lagged items, trading volume, and volatility to establish a vector autoregression model to predict the trend of the Shanghai Composite Index;

In the risk assessment module, the system divides the risk into five levels, the first level is extremely low risk, the second level is low risk, the third level is medium risk, the fourth level is medium-high risk, and the fifth level is high risk, which prompts the stock market as a whole risk.

The result output module outputs the overall risk level of the stock market and prompts the risk. The fifth level of high risk is suitable for aggressive investors, the fourth level of medium and high risk is suitable for active investors, the third level of medium risk is suitable for balanced investors, and the second level of low risk is suitable for investors. Steady investors, level one lower risk is suitable for conservative investors. The risk level of the stock market can not only serve investors' decision-making, but also provide a basis for the government to formulate relevant policies and enterprises to implement corresponding strategies.

In the above-mentioned embodiment of the present invention, a text mining method is provided, as shown in FIG. 3 ,

Data sources include policy news, financial news, and forum data. Sources of policy news include the China Securities Regulatory Commission, China Banking Regulatory Commission, the Central Bank, and news broadcasts. Financial news sources include Hexun.com and Oriental Fortune. Forum data sources are Sina Finance Forum, NetEase Finance Forum and Tencent Finance Forum. Conduct text analysis processing on the above news sources to mine market sentiment and investor sentiment;

1), text word segmentation, apply the word segmentation system to process word segmentation of text data;

2), part-of-speech tagging, after removing stop words, modal particles, etc., tag the words with part-of-speech;

3) Emotional polarity labeling, emotional polarity labeling of words, divided into positive words, negative words and neutral words, and counting the number of positive words and negative words respectively;

4) Calculation of emotional index, according to the emotional calculation formula (1), the emotional index of each news or forum comment data can be obtained, so as to obtain the daily emotional index, where Sdx represents the emotional index, Nn represents the number of negative words, Np The number of positive words, the sentiment index greater than 0 represents pessimistic investor sentiment, and the sentiment index is less than 0 represents optimistic investor sentiment;

5) Sentiment index adjustment. In Step 104, it is found that news on government websites is special. Policy news is influential within a certain period of time and policy news is sparse. That is, the absence of policy news does not mean that the government has no emotional expression, but The emergence of policy news represents the attitude of the relevant regulatory authorities towards the stock market for a period of time, so the time decay factor is set to adjust the policy news. The adjusted policy news index is represented by , and the calculation formula is shown in (2), Denotes the ith (i=0,1,2) period lag item of the original policy news index, where

is a monotonically decreasing time decay function, and the calculation formula is shown in (3);

6) Sentiment index integration. Combining the sentiment indexes of 104 and 105, the policy sentiment daily index, financial sentiment daily index, forum sentiment daily index and comprehensive sentiment daily index can be obtained.

The present invention is not limited to the preferred embodiment described above, and anyone should know that any structural changes made under the inspiration of the present invention, and any technical solutions that are the same as or similar to the present invention, all belong to the protection scope of the present invention.

Claims (2)

1. stock market's risk profile platform, it is characterised in that:

Including:

Data acquisition module, for automatically collecting and obtaining stock market transaction data and multi-source internet text notebook data；

The data obtained in data acquisition module are carried out pretreatment, comprise data cleansing, data set by data preprocessing module Become, data convert and data regularization, carry out Data Preparation for setting up Stock Market Forecast Model；

Text mining module, is used for being analyzed internet text notebook data processing to excavate investor sentiment, builds emotion and refer to Number, comprises text participle, part-of-speech tagging, feeling polarities mark, moos index calculating, moos index adjustment, moos index integration Six big steps；

Stock Market Forecasting module, stock market is predicted point by integrated application text mining, machine learning, the method for mathematical statistics Analysis；

Risk evaluation module, carries out risk according to the result of Stock Market Forecasting module to stock and the market overall trend of monitoring in real time Grade classification；

Result output module, the risk class of the stock for being paid close attention to investor output, and export whole market simultaneously Risk class situation also provides real-time early warning.

The text mining method of stock market the most according to claim 1 risk profile platform, it is characterised in that:

Text mining method is that a kind of method of structural data that is converted into by non-structured text data is to analyze in document Viewpoint, attitude or the emotion contained；

The internet text database that text mining method is used comprises POLICY, financial and economic news, forum data three aspect, POLICY can excavate attitude and the tendency of government, and financial and economic news it will be seen that socioeconomic integrated information, forum data Can the most directly extract investor sentiment；

Text mining module in stock market's risk profile platform is that the text data in the Internet is entered by applicating text method for digging Row analyzing and processing, thus extracts the viewpoint of investor, attitude, emotion, then using the moos index calculated as input Variable is applied in Stock Market Forecasting module.