[pt] Investidores estão sempre à procura de uma vantagem. Porém, tradicionais teorias financeiras nos dizem que tentar predizer tendências na bolsa de valores é um esforço em vão, uma vez que seguem um passeio aleatório, i.e., um processo estocástico ou randômico. Além disso, afirma-se que o mercado é eficiente de maneira que sempre incorpora e reflete toda informação relevante, o que torna impossível bater o mercado. Recentemente, com o crescimento da web e aumento da disponibilidade de dados em conjunto com a evolução dos algoritmos de Aprendizado de Máquina, diversos trabalhos tem aplicado técnicas de Processamento de Linguagem Natural em notícias financeiras e dados de redes sociais para prever variações do preço de ações. Consequentemente, estão surgindo fortes evidências que o mercado pode, em algum grau, ser previsto. Este trabalho descreve o desenvolvimento de uma aplicação baseada em Aprendizado de Máquina para realizar a predição de tendências no mercado de ações, i.e., variações negativas, positivas ou neutras de preços com granularidade de minuto. Avaliamos o sistema usando dados de cotação de ações da B3 (Brasil Bolsa Balcão), antiga BM&FBOVESPA, e um dataset de tópicos mais relevantes buscados no Google Search e seus artigos relacionados, que são disponibilizados pela plataforma Google Trends e coletados, minuto a minuto, de 15/08/2016 até 10/07/2017. Os experimentos mostram que esses dados provêem informação relevante para a tarefa em questão, onde conseguimos uma acurácia de 69.24 porcento para a predição de tendências do ativo PETR4, criando alguma / [en] Investors are always looking for an edge. However, traditional economic theories tell us that trying to predict short-term stock price movements is wasted effort, since it approximate a random walk, i.e., a stochastic or random process. Besides, these theories state that the market is efficient enough to always incorporate and reflect all relevant information, making it impossible to beat the market. In recent years, with the growth of the web and data availability in conjunction with advances in Machine Learning, a number of works are using Natural Language Processing to predict share price variations based on financial news and social networks data. Therefore, strong evidences are surfacing that the market can, in some level, be predicted. This work describes the development of an application based on Machine Learning to predict trends in the stock market, i.e., positive, negative or neutral price variations with minute granularity. We evaluate our system using B3 (Brasil Bolsa Balcão), formerly BM&FBOVESPA, stock quotes data, and a dataset with the most relevant topics of Google Search and its related articles, provided by the Google Trends platform and collected, minute by minute, from 08/15/2016 to 07/10/2017. The experiments show that this data provides useful information to the task at hand, in which we achieve 69.24 per cent accuracy predicting trends for the PETR4 stock, creating some leverage to make profits possible with intraday trading.

[pt] Neste trabalho, estudo o comportamento dos arbitradores ao se depararem com variações nos níveis de preços das ações, mais especificamente, analisando a forma como eles buscam informações sobre esses ativos. Para isso, testo e confirmo a hipótese de que os retornos anormais das ações levam os investidores a buscarem ativamente mais informações sobre essas empresas, usando dados de volume de buscas no Google. Adicionalmente, analiso de forma separada o impacto de retornos anormais negativos e de retornos anormais positivos sobre o volume de buscas do Google, chegando à conclusão de que os retornos negativos têm um efeito maior sobre o volume de buscas que os efeitos positivos. / [en] I study the behavior of arbitrageurs when they are faced with changes in stock price levels, more specifically analyzing the way they seek information about these assets. I test and confirm the hypothesis that abnormal stock returns prompt investors to seek actively information about these companies by using Google search volume data. Furthermore, I study the separate effects of negative abnormal returns and positive abnormal returns on Google search volumes, and conclude that negative returns cause a greater impact on the search volumes than positive returns.

In this paper, I analyze Bitcoin price formation and adoption rates at a global and national level. In determining Bitcoin prices, I consider contemporaneous and lagged values of traditional determinants of currencies, such as inflation and industrial production, and digital currency specific factors, primarily public interest. Using monthly time-series data across five years (2011 – 2016), I find that global public interest in Bitcoin, measured by Google searches for the keyword ‘Bitcoin,’ has a positive and significant impact on Bitcoin prices. I extend the analysis to a country level by employing a proxy for adoption rates, represented by the number of local Bitcoin client downloads, which is a useful predictor of prices. I examine pooled data across 12 countries to show that searches for ‘Bitcoin’ can be used to predict adoption rates and, consequently, prices. To the best of my knowledge, this is the first academic article to study Bitcoin usage at a national level. I find that contemporaneous values of traditionally used macroeconomic determinants of currency prices, except inflation, do not have a significant impact on Bitcoin prices.

Dissertação (mestrado)—Universidade de Brasília, Faculdade de Comunicação, Programa de Pós-Graduação em Comunicação, 2018. / Este trabalho se propõe a identificar o que os internautas que utilizaram o Google no Brasil entre 2004 e 2014 expressaram no motor de busca sobre a criatividade. Para isso, extraímos os 50 termos de busca mais frequentes com essas características na ferramenta Google Trends e fizemos dois levantamentos bibliográficos dos trabalhos científicos que utilizam essa mesma metodologia: um sobre os textos produzidos no campo da comunicação social e o outro no campo da criatividade. A metodologia também contemplou a investigação dos termos de busca por meio do método da Análise Temática (AT) para criar o Mapa da Imagem da Criatividade. Essa representação gráfica possibilitou a verificação das afinidades e discrepâncias dos 50 termos com a teoria acadêmica a respeito da criatividade e a compreensão dos seus níveis de comunicabilidade. A criação do mapa também possibilitou a percepção de parte do imaginário da criatividade expressa nos termos analisados. / This work intends to identify the expression of the image that the internauts who used Google in Brazil between 2004 and 2014 have about creativity. We extracted the 50 most frequent search terms with these characteristics in the Google Trends tool and carried out two bibliographical surveys of the scientific works that use the same methodology: one on texts produced in the field of social communication and the other in the field of creativity. The methodology also contemplated the investigation of search terms through the method of Thematic Dialogic Analysis (AT) and discourse analysis to create the Image Map of Creativity. This graphical representation made it possible to verify the affinities and discrepancies of the 50 terms with the academic theory regarding creativity and the comprehension of their levels of communicability. The creation of the Map also allowed the perception of part of the imagery of the creativity expressed in the analyzed terms.

Diploma thesis refers about effect of informational cascade, which is causing herding behaviour, in sector indices. Thesis distinguishes between market and sector informational cascade. Each of them is represented as an indicator made from dataset provided by Google trends service. Effect is demonstrated on extended version of CAPM theoretical concept to multi factorial model including the indicators, which is based on APT theoretical concept. For the purpose of robust analyses is then realized Granger causality on regression results.

In recent years, modern prediction models have evolved to include behavioural data such as user-generated search query data that capture market sentiment and reach beyond the grasp of established macroeconomic indicators. These applications had considerable success in predicting a wide range of economic phenomena with the assumption that internet interaction behaviour resembles probable offline behaviour. Despite the considerable success of this approach, the existing literature argues for the continuous validation of search query keywords and its probable meaning over time to avoid spurious and biased results. Although recent literature attempted to bridge the keyword validation gap, this line of research is still in its infancy. This thesis sets out to examine the validity of web search intention to serve as a “pure” demand proxy for direct real estate market prediction in the United Kingdom. More specifically, it is directed towards constructing web search indices to explore: (i) the extent to which an individual’s true real estate orientated intentions manifest themselves in their web search behaviour and (ii) the magnitude to which real-time information adds value towards the prediction of illiquid asset classes. In doing so, a conceptual framework is produced, which outlines the logic and importance associated with intention specific web search in the digital age, as well as its relation to real estate demand. The empirical findings suggest that intention specific keyword development might be of little importance for aggregate housing and office market forecasts in the United Kingdom. On the contrary, it seems that the viability of intention specific web search keyword development increases when it is directed at a specific regional market. The overall thesis narrative introduces a new way of thinking about web search in the context of economic demand and draws from a variety of principles and methodologies to establish an avenue from which future research can be conducted.

No description available.

The 2018 Swedish general election saw the largest collective polling error so far in the twenty-first century. As in most other advanced democracies Swedish pollsters have faced extensive challenges in the form of declining response rates. To deal with this problem a new method based on search query data is proposed. This thesis predicts the Swedish general election using Google Trends data by introducing three models based on the assumption, that during the pre-election period actual voters of one party are searching for that party on Google. The results indicate that a model that exploits information about searches close to the election is in general a good predictor. However, I argue that this has more to do with the underlying weight this model is based on and little to do with Google Trends data. However, more analysis needs to be done before any direct conclusion, about the use of search query data in election prediction, can be drawn.

Stejskalová, J. Impact of the information on changes in tax burden on the stock market. Diploma thesis. Brno: Mendel University, 2017. The thesis investigates the relationship between the stock price returns and news about the tax burden of US companies listed on NASDAQ. Special emphasis is put on the role of perception of the news related to changes in tax burden. Using application Google trends, we show that increasing tax searches decrease stock prices. The thesis also investigates the positive relationship between news about tax burden and stock prices, in particular, shocks. Additionally, we differentiate between the market capitalization using interactions with the dummy variables. The results confirmed a higher impact of perception on large cap companies, we point out the importance of sentiment analysis at liquid markets.

Today’s competitive market in the video game industry puts a lot of stress on companies to be ahead of competitors. The ability to predict the potential of a product gives companies an advantage over competitors on the market. Companies have therefore increased Competitive Intelligence (CI) departments in recent years and looked for ways to optimise forecasting capabilities. Researchers argue for the use of Machine Learning (ML) to forecast market potential of products, and have investigated varying methods of optimising the accuracy of models. Past studies have shown the existence of predictive value in online search traffic on Google. This study set out to investigate if Youtube search traffic holds similar predictive value. Results show that Youtube trends do have a degree of inherent predictive value, and the addition of the information enhances forecasting performance of ML models. However, the exact degree of the predictive value in Youtube trends is yet to be determined, as some evidence from testing implicated it to be strong while others weak.