Abordagens computacionais para Big Data aplicadas ao mercado de ações: um mapeamento sistemático

Abstract

Context: One of the biggest problems in the decision-making process in organizations is the low rate of knowledge reuse. It this occurs because most of the knowledge in organizations is not addressed, making articulation difficult. So, in this context the Knowledge Management (KM) emerges as an important means to manage useful knowledge. Knowledge generated from discussions can become a valuable item. In the financial sector, for example, it can help investors in their decision-making about the actions of certain companies, which are traded on a stock exchange, which at a given moment are more attractive. In this sector, the data volume, the speed of generation and data process of from different sources create challenges, such as storage, processing, visualization and, mainly, data analysis. The large data volume produced by different distributed and decentralized sources that quickly generate data with complex and evolving relationships is called Big Data. The analysis of this large data volume aims to extract information about the domain and the result of this analysis can help organizations in decision making. Goal: The objective of this work is to summarize the main approaches of Big Data applied in the financial market in order to identify the state of the art in the area, as well as future research. Method: An systematic mapping was conducted. A systematic mapping provides a broad view of a research area to determine if there is research evidence about a particular topic. Results: 115 relevant studies were selected to answer the research questions. The results showed that the research topic is recent and has a wide variety of approaches to Big Data applied to the stock market. Sentimental analysis and the use of time series, for example, are commonly applied across studies. Conclusion: According to the results obtained, it is possible to affirm that researchers are investing in the creation of new tools, techniques, algorithms and models to deal with large sets of real-time and historical stock market data. Models based on artificial neural networks and machine learning algorithms are recurrent in research. It was found that it is possible to create approaches for forecast analysis in the stock market using data sources of different natures, such as social media and historical price reports. It is hoped that the results achieved by this research can provide guidance to appropriately position new research activities on the investigated topic.