Public Opinion Monitoring And Analysis Platform For Financial Market

The scale of the financial market has been expanding with the development of the economy and the improvement of people’s expectations for life.In recent years,the number of investors pouring into the financial market has gradually increased.For investors,investing in the financial market at this stage faces huge potential opportunities,and investors also need to avoid the risks that accompany the opportunities as much as possible.By mastering the public opinion status of the financial entities concerned in the financial market,combining the financial public opinion and the price trend of the financial entities,investors can make decisions as soon as possible and make trading strategies,which will help investors avoid risks in time and make profits in the financial market.Therefore,combining with natural language processing technology,this paper proposes a public opinion monitoring and analysis technology for financial entities in the financial market to help investors avoid risks and seize opportunities.The financial market public opinion monitoring and analysis technology has the following difficulties:First,identify financial entities in massive financial texts and extract financial texts related to target financial entities.At present,there are few named entity recognition models applied in the financial field,and the size of the named entity recognition data set in the financial field is small,resulting in the model being unable to fully extract features,thus affecting the classification performance;Moreover,at this stage,there is no financial public opinion processing model that integrates the financial market text.Although some of the existing work has achieved text screening for a single financial entity,it only represents the characteristics of financial market public opinion through the emotional classification of the financial news text or the financial community public opinion text and only makes a simple accumulation operation,without considering the different impact of the events represented by different financial texts on the financial market.And this public opinion analysis method is based on the condition that a single financial entity is not affected by other financial entities in the financial market,while there is no independent financial entity in the real financial market that is not affected.The evaluation of public opinion of financial entities needs to consider the impact of other financial entities,so this paper believes that a method to measure the degree of correlation between financial entities is needed.Aiming at the above problems,this paper implements a named entity recognition model based on multi-feature fusion for the financial field,and a financial entity public opinion analysis model based on text weight and co-occurrence matrix.On this basis,it builds a public opinion monitoring and analysis platform for the financial market.The main work of this paper is as follows:In this paper,a named entity recognition model based on multi-feature fusion is introduced.Through the introduction of image-based font features and phonetic-based character features,combined with the semantic features obtained from the preprocessing language model,the influence of the number of features and the fusion method on the performance of the model is explored on four data sets.The experiment shows that the method proposed in this paper can enhance the character feature information.This paper applies the model that has made a performance breakthrough in the named entity recognition dataset in the small-scale financial field to identify financial entities and determine the entity related text.On the basis of filtering a single financial entity and its related text based on the named entity recognition task,this paper proposes a complete process of financial public opinion analysis combined with text weight:according to the type of financial text,the financial text is divided into industry policy text,company announcement text,financial news text,community public opinion text and other levels,and a text training set format combined with market status labeling is proposed;A public opinion analysis model based on multiple convolutions and attention is built.Different types of text features of the day are extracted through multiple convolutions and the weight is learned through the attention mechanism to generate a feature vector that integrates the public opinion of the day,and finally the public opinion results of the day are obtained;The co-occurrence matrix of financial entities is also introduced to assist in the evaluation of public opinion status.In this paper,the classification accuracy and cointegration coefficient are used to verify the scientificity and effectiveness of this method.Supported by the realized data extraction framework and with model and algorithm as the core,this paper builds a public opinion monitoring and analysis platform for the financial market,which is mainly based on the functional interface of financial data analysis and the large visual screen.