Research And Implementation Of Text Mining System For Big Data Accurate Investment Promotion

Investment promotion is one of the important means to optimize the industrial structure and stimulate regional economic growth.As the Internet development enters the era of big data,the methods of screening investment companies through traditional manual methods are inefficient and the analysis results are not comprehensive.Therefore,in the process of government investment promotion,the need for combining big data technology with investment promotion work to achieve precise investment promotion is becoming stronger.At present,the existing investment management system can perform basic information processing on the company's registration information,investment amount and other related data,but how to analyze the large amount of data collected on the Internet has attracted more attention.In this thesis,by using big data text mining technology to analyze corporate financial public opinion and patent technology achievements,and obtains the current status of corporate financial public opinion and companies with advanced patent results.it helps investment personnel to accurately assess enterprise risks,narrow the scope of target investment enterprises,and improve the accuracy of investment promotion.In the evaluation of corporate financial news and public opinion,the combined model of ALBERT and CNN is used to analyze the sentiment of the company's annual report text to obtain the market's emotional attitude to the corporate financial status.The model uses ALBERT to extract emotional features and train word vectors in the company's annual report,and then input the word vectors into CNN to train the emotion classifier.The study showed that this model compared to the word vector Word2 Vec model is trained as an original feature input,the effect on sentiment analysis has improved significantly,indicating that the use of ALBERT-CNN sentiment analysis model is a combination of financial results of public opinion tendency is feasible.The industry's patented technical achievements are analyzed using a multi-level patent text classification combined model that integrates ALBERT and Bi GRU.The ALBERT model improves the characterization ability of word vectors,and uses Bi GRU to preserve the semantic association between long-distance words in the patent text to the greatest extent,and improves the effect of text classification.The respective experiments performed well,indicating that it is feasible to use the ALBERT-Bi GRU combined model to automatically classify patents.According to the actual needs of investment promotion staff for accurate investment,the analysis and design of a text mining system for big data and accurate investment has been realized.The system implements the query of the basic registration information of the enterprise through four parts: enterprise information management,enterprise public opinion evaluation,patent technology management,and user management;sentiment analysis of the company's financial annual report and visual display of the propensity of the company's public opinion;automatic classification of patent texts to facilitate the inquiry and management of investment personnel;as well as the information modification and authority management of system users.The experimental results show that the text mining system designed in this thesis can effectively complete the analysis of related texts in the process of investment promotion,and support the investment personnel to view the status quo of corporate public opinion in the system and query related corporate patent information.Reduce the work burden of investment promotion personnel and help them to screen investment enterprises,thereby improving the accuracy of investment promotion work.