Analyzing Almost 10 Million Tweets, Research Finds Public Mood Can Predict Dow Days in Advance
IU News Room (10/18/10) Steve Chaplin
Indiana University researchers found that by analyzing millions of tweets they can predict the movement of the Dow Jones Industrial Average (DJIA) up to a week in advance with near 90 percent accuracy. Indiana professor Johan Bollen and Ph.D. candidate Huina Mao used two mood-tracking tools to analyze the text content of more than 9.8 million Twitter feeds and compared the public mood to the DJIA's closing value. One tool, called OpinionFinder, analyzed the tweets to give a positive or negative daily time series of public mood. The other tool, Google-Profile of Mood States, measures the mood of tweets in six dimensions--calm, alert, sure, vital, kind, and happy. The two tools gave the researchers seven public mood time series that could be matched against a similar daily time series for the DJIA closing. "What we found was an accuracy of 87.6 percent in predicting the daily up and down changes in the closing values of the Dow Jones Industrial Average," Bollen says. The researchers demonstrated that public mood can significantly improve the accuracy of the basic models currently used to predict Dow Jones closing values by implementing a prediction model called a Self-Organizing Fuzzy Neural Network.
Monday, October 18, 2010
Blog: Analyzing Almost 10 Million Tweets, Research Finds Public Mood Can Predict Dow Days in Advance
Thursday, September 9, 2010
Blog: Most Influential Tweeters of All
Most Influential Tweeters of All
Northwestern University News Center (IL) (09/09/10) Erin White
Northwestern University researchers have designed a Web site that tracks the top trending topics on Twitter in real time. The Web site--pulseofthetweeters.com--uses an algorithm to rank the most influential people tweeting on trending topics. The researchers say the algorithm combines dynamic data mining, sentiment analysis, and network analysis in real time. In addition to identifying the most influential tweeters, the algorithm can tell users whether the tweets are positive, negative, or neutral. "Discovering patterns, opinions, and sentiments from massive number of tweets is challenging in itself, but discovering influencers and leaders for specific topics is a major technological advance in data mining," says Northwestern professor Alok Choudhary. "The good thing about our system is it's completely automatic, and it needs minimal human supervision," says Northwestern's Ramanathan Narayanan.
Monday, June 21, 2010
Blog: Blogs and Tweets Could Predict the Future
Blogs and Tweets Could Predict the Future
New Scientist (06/21/10) Giles, Jim
Forecasts about social and economic trends could be generated through the analysis of blogs and tweets, building on earlier research by Google and others to mine the frequency of specific search terms to outline purchasing patterns. With blogs and tweets added to the equation, trends other than buying behavior--such as political sentiment and stock market patterns--could possibly be predicted. For instance, researchers at the University of Illinois at Urbana-Champaign were able to forecast stock market behavior by using more than 20 million blog posts to build an "Anxiety Index" that measures the frequency with which a range of words associated with apprehension, such as "nervous," show up in the posts. The appearance of these terms correlated with lower stock prices. Tools that quantify the national mood could prove useful to stock traders, who will be more likely to refrain from taking risks if they know consumers are fraught with pessimism, for example. Researchers say that Web data analysis methods could be used to make even more accurate predictions as researchers devise more refined techniques for measuring the emotional content of blogs and tweets.
Tuesday, May 11, 2010
Blog: Carnegie Mellon Study of Twitter Sentiments Yields Results Similar to Public Opinion Polls
Carnegie Mellon Study of Twitter Sentiments Yields Results Similar to Public Opinion Polls
Carnegie Mellon News (05/11/10) Spice, Byron
Carnegie Mellon University (CMU) researchers analyzed the sentiments expressed in a billion Twitter messages during 2008-2009 relating to consumer confidence and presidential job approval ratings and found that they were similar to those of well-established public opinion polls. The research suggests that studying tweets could become an inexpensive, rapid way of gauging public opinion on some subjects, says CMU professor Noah Smith. "With seven million or more messages being tweeted each day, this data stream potentially allows us to take the temperature of the population very quickly," Smith says. The Twitter-derived sentiment measurements were much more volatile day-to-day than the polling data, but when the researchers looked at the results over a period of days, they often correlated closely with the polling data. The researchers say that improved natural-language processing tools, query-driven analysis, and the use of demographic and time stamp data could enhance the sophistication and reliability of the measurements.
Monday, May 3, 2010
Blog: Computer Science Shows How 'Twitter-Bombs' Wield Influence
Computer Science Shows How 'Twitter-Bombs' Wield Influence
Wellesley College (05/03/10) Corday, Arlie
Wellesley College computer science professor P. Takis Metaxas says "Twitter bombs"--sending many Tweets from a large number of Twitter accounts within a short period of time--are being used to affect the outcome of elections. Metaxas says Twitter bombs were used against U.S. Senate candidate Martha Coakley in the recent Massachusetts senatorial election. A Twitter bomb reaches many people very quickly. "In addition, because Google is displaying Twitter trends in a prominent place, you influence Google search results," Metaxas says. The result of the Twitter bomb was "disproportionate exposure to personal opinions, fabricated content, unverified events, lies, and misrepresentations that would otherwise not find their way in the first page (of Google search results), giving them the opportunity to spread virally," he says. In an analysis of the Coakley Twitter bomb, the researchers found that the attack was launched by the American Future Fund, the same group that attacked John Kerry's record during his 2004 presidential campaign. Metaxas is developing software to detect Twitter bombs in real time.