Exploratory Research On Topic Investment Based On Text Mining

With the vigorous development of the Internet and social media,text information on social platforms has subtly influenced investors' investment behavior,and market public opinion directly affects investors' psychology and behavior.Therefore,the rotation of investment topic style is an important factor for market driving.Using the massive amount of Internet text information data generated by investors to accurately give investment advice on topic-related concept stocks will have great practical significance for the theoretical exploration and practical operation of investors' investment analysis models.In this paper,through in-depth analysis and mining of the stock bar text,for the first time in the field of quantitative investment,an investment strategy model based on the combination of topic mining and sentiment analysis is constructed.The method is mainly based on the Latent Dirichlet Analysis(LDA)algorithm model,and the topic mining of the text of Dongfang Fortune Finance Financial Review Bar is implemented to realize the daily public opinion hotspot monitoring of the stock market;then the information entropy is designed to improve the uniqueness of the topic mining results.The indicator further filters the topic interference words.This method effectively reduces the confusion of the topics and improves the difference between the topics;the topics mined by the topic model are random and do not match the topics recognized by the financial market.Existing concepts in the market-new energy vehicle topic as an example,using the optimized topic model to extract keywords of the topic to build a thesaurus,designing a "topic score" indicator and then extracting the popularity of the topic from the public opinion of the stock bar and implement automatic monitoring every day to complete the quantification of topic heat.Next,this article builds a quantitative investment strategy based on this and conducts an empirical analysis: First,the public opinion mining of the stock bar during 2019/01/01-2020/02/29 extracts the topic score of new energy vehicles,and compares it with the new energy vehicle index excess income.The volatility was tested for correlation,and the topic-related stocks that were significantly related to the topic score were selected as the investment stock pool.In this regard,I comment on the stocks of the stock pool to crawl the stocks of the same period,and use the text classification algorithm for sentiment analysis to construct a "bull" sentiment indicator to assist in the generation of strategic timing signals;finally,the combination of "topic score" + "bull indicator" and the moving average strategy establishes a new quantitative investment strategy model based on the "topic + sentiment" of the stock bar text.After a historical backtest on the Join Quant platform,it confirmed that the strategy is effective and performs well,thus providing investors with a new investment strategy.The model's well performance also proves that the stock bar text does contain valuable information for thematic investment decisions.