Since its founding in 2009, crowdfunding platform Kickstarter has hosted a total of 118,422 projects-- with a little over 50,000 reaching their funding goals. Over the last few years, a few enterprising users have even turned their successfully-funded projects into full-fledged businesses (think Pebble or OUYA). The site also sees a fair amount of projects fail each year.

This begs the question: How do you know which projects will succeed?

A Swiss research team recently developed a tool that they claim can predict which Kickstarter projects will succeed and which will fail-- with 76 percent accuracy within four hours of launch. 

In the research paper Launch Hard or Go Home! a three person research team from the university École Polytechnique Fédérale de Lausanne outlined their method of predicting the success of individual Kickstarter projects using a combination of data mining and analysis of social media and Kickstarter activity.

The team scraped data from Kickstarter’s website between September 2012 and May 2013--covering a total of 16,042 campaigns backed by 1,309,295 users. From the collected data, the research team built a prediction model based on the flow of pledged money, the number of backers and the buzz surrounding the project on Twitter.

The model found that there was a strong correlation between the number of tweets about a project and the project’s success (successful campaigns averaged 73 tweets, while unsuccessful campaigns averaged only 20). They also found that projects with backers who had backed at least one other campaign were also more likely to succeed. 

This is not the first time that people have tried to predict the success of campaigns on Kickstarter, however. There have been other models developed in the past such as Kicktraq and CanHeKickIt

Kickstarter declined to comment on the story.