Sentiment analysis applied to decision support in financial market

• Grant number: 18/24371-1
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: February 01, 2019
• End date: September 30, 2019
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Guilherme Palermo Coelho
• Grantee: Brenda Alexsandra Januário
• Host Institution: Faculdade de Tecnologia (FT). Universidade Estadual de Campinas (UNICAMP). Limeira , SP, Brazil

Abstract

Investment in the stock market has grown in Brazil in the last two years, especially when taking into account the number of individual investors. According to data from September 2018, the growth in this type of investor is already at 18%, compared to the end of 2017. Given this scenario, it is increasingly important to better understand the oscillations of the stock market and develop decision support systems for this type of investment. Thus, the present work seeks to contribute in this sense by both exploring automatic sentiment analysis in news related to companies listed on São Paulo's Stock Exchange (BM&FBovespa) and verifying the existence of correlations between the average sentiment identified in the news associated with a given company, in a given period, and the values of its shares in the stock exchange. For this, the following activities will be developed here: (I) construction of a news database related to a company listed on BM&FBovespa, manually labelled with the sentiments associated with each news; (II) adaptation of a sentiment lexon, aimed at the financial market domain, to Portuguese; (III) correlation analysis between the average sentiment observed in the collected news and the closing values of the company's stocks; and (IV) application of two automatic sentiment analysis strategies to the created database, one based on text mining and another on machine learning, and comparison of results. This project is directly related to a PhD research currently under development in the HighPIDS research group (http://highpids.ft.unicamp.br).