Financial Forum Text Mining Based On Topic Model

The financial forum,as a platform for shareholders,is very popular.Making good use of the text information in the financial forum is of great help to investment.For the netizens who visit the financial forum,the most wanted information is mainly two: recent events with high degree of discussion and what do netizens think about stocks.This thesis will mine these two kinds of information through the topic model.The main work of this thesis is as follows:Mining the topic of the news section.The replys in forum corpus are always short,and there are a lot of water posts in the forum.Traditional topic models are not suitable for short texts and do not consider meaningless texts,so it's hard to get good results.In view of these characteristics of the forum corpus,this paper proposes the BBS-LDA topic model.This model samples the topic in terms of sentences,and the sentences in each post have the same topic distribution.This approach considers the structural characteristics of the forum and can alleviate the sparsity problem.At the same time,the model introduces meaningless topics and user information to alleviate the impact of water posts on topic mining.Through a comparative experiment with real corpus,the model can improve the quality of keywords.Emotional Analysis of stock section.Forum corpus has no annotated information,and many supervised classification methods need manual annotation,so they are not applicable.This thesis constructs a financial sentiment dictionary through Word2 vec and SO-PMI,and uses this as a supervised information to use JST(Joint sentiment topic model)to analyze the sentiment of stock section.The effectiveness of the method is demonstrated by experiments on manually labeled data sets.Finally,based on the above research,the financial forum text analysis system was developed.The system can automatically crawl the text information of financial forum through the crawler and display it to users after the algorithm processing.Besides,the system also provides some interfaces for editors to adjust topic weights in order to improve the practicability of the system.The development of the system can make it easier for netizens to obtain financial forum information more intuitively and quickly.