Twitter’s unique design allows it to easily capture real-time, location-specific insights and sentiments about anything from current events to emerging trends to economic data. By analyzing public tweets for specific keywords, forecasting tools with a high level of accuracy can be (and have been) developed.

This article discusses some real world examples of how tweets have been used to make predictions.

Yes! Twitter Can Tell The Future…

Three days before the official announcement of Ebola cases in West Africa, tweets about the outbreak already reached over 60 million users. This is according to a study that appeared in a 2015 issue of the American Journal of Infection Control. Twitter is a uniquely powerful social media platform because it is designed to capture real-time sentiments and insights on current events and emerging trends. In fact, researchers have long been harnessing Twitter data’s rare forecasting power for stock movements, influenza outbreaks, etc.

Here are some notable examples of how Twitter is being used as a forecasting tool.

Forecasting Economic Data Using Twitter

Researchers at the University of Michigan and Stanford University believed (and have proved) that Twitter can add insight to economic predictions. For some years now, these computer scientists and economists have been forecasting initial jobless claims by doing an analysis of public tweets. The project is the “University of Michigan Economic Indicators From Social Media” and it has been running for about 3 years as of this writing.

Their method scans billions of tweets for terms related to job changes, and analyzes them. They then attempt to predict initial jobless claims using the data from the tweets.

While Twitter-based predictions is not intended to replace predictions from economists, the idea is that combining both could produce very accurate forecasts.

When you think of how powerful social media could become in the future, it is exciting to imagine that we could potentially start trading the stock markets with social media data as our primary indicator!

In fact, trading the stock markets using Twitter data is already being done (or experimented with) by the guys at TweetTrader.net.

Coronary Heart Disease Rates

The paper “Psychological Language on Twitter Predicts County-Level Heart Disease Mortality” in a 2015 issue of the journal Psychological Science detailed important findings on analyzing publicly available, location-specific tweets in 2009 and 2010. It was discovered that the prevalence of negative emotional topics and word choices (like expletives and the word “hate”) in a certain county is strongly tied to the mortality rate for heart disease in that area. The study made use of Twitter analysis cross-checked with county-level health data, including smoking rates and obesity levels. The findings of this study reinforced the existing sociological research that suggests that the combined characteristics of communities can be more predictive of physical health than the reports of any one individual.

Influenza Forecasting

According to this paper, Twitter data could help track and predict flu outbreaks. By scanning and analyzing tweets, flu forecasting can be significantly improved and error seen with standard prediction models that use data from the Centers for Disease Control and Prevention can be reduced by 17 to 30 percent. Twitter’s data is also fast and readily available, whereas data from the CDCs is often delayed by a week or two.

Probably the most famous example of using big data for public health predictions to date is Google Flu Trends. Google Flu Trends is based on key word searches on Google and since its launch in 2008, its accuracy has been challenged. It overestimated the instance of flu in 2012/2013 and underestimated the swine flu of 2009.

Since Twitter data isn’t based on keyword searches, it could potentially produce more accurate results than Google Flu Trends. Mark Dredze (an author on the PLOS Currents paper mentioned above), believes that when combined with the CDC’s data, Twitter data predictions are more accurate than CDC predictions alone or models that use Google’s keyword searches.

Suicide Rates

The paper “Tracking Suicide Risk Factors Through Twitter in the US” which appeared in an issue of the journal Crisis [2014;35(1):51-9. doi: 10.1027/0227-5910/a000234], demonstrated that states registering high numbers of public tweets about suicide and bullying have high actual suicide rates. This forecasting capability of Twitter may be tapped more efficiently by systematically identifying high-risk states and intensifying the efforts of suicide crisis centers in those flagged areas.

Conclusion

These are just a few examples of how social media and Twitter particularly can be used to forecast the future. Some other examples where the Twitter method has been (or is being) employed to make predictions include:

• Predicting Elections
• Traffic at e-commerce sites (as a proxy for retail sales)
• Movie box-office sales
• Mortgage refinancing
• Gas prices