Sentiment Analysis on News Articles Application for Stock Returns Prediction

Kotelnikova Daria, 1-st year MSc, ICEF HSE

Review on research presentation

 “Sentiment analysis on news articles application for stock returns prediction”

 of Alexey Savushkin

Basically, traders get information from social networks and newspapers for predicting stock prices, because news are believed to have impacts on stock price return (which appears to be not a new idea, as articles, concerning this matter, have been written since 2014 [1]).  Microblogging today has become a very popular communication tool among Internet users. Millions of users share opinions on different aspects of life every day. Therefore, microblogging web-sites, such as Twitter, are rich sources of data for opinion mining and sentiment analysis. So, tools, which could help to analyze data in real time from these sources, will be very useful.

In my opinion, this topic is highly relevant, because analyzing information and to understand the overall mood of the society, concerning different matter, would help traders to gather information and to make their predictions more precise.  So, the motivation of the research is clear and up-to-date, but the idea is not new, as it’s talked a lot about and a lot of research has been made on this matter. (quick search in Google.scholar and Scopus give more than 223.000 results).

Also, it’s sure, that the relationship between financial markets and “mood” of news exists and there are a lot of examples., some of them were shown in the presentation.

The author emphasizes, that the contribution he makes in the research in the following:

• New sources of data (not necessary by Twitter API),

• Applying modern algorithms of extracting data and sentiment analysis, which haven’t been used in the other papers on the subject.

Methodology is the following:

• Choosing opinion leaders - the most read authors by topic in the last days (Donald Trump, Elon Musk and etc. ~ 50 people), choosing social networks – Twitter, choosing newspapers – Financial Times, WSJ
• Developing algorithm and extract data from articles by methods of parsing in python – by Beautiful Soup, choosing algorithm of sentiment analysis
• Checking causality between returns of S&P 500 stock returns and normalized difference between positive and negative news by Granger causality test

As a result, author wants to see, whether the social activity in «financial part» of Internet indeed influences stock price returns.

Speaking of the comments on the thesis, it’s not really clear, why the interval of 100 days has been chosen for analysis, is it random or are there some assumptions?