Bing Predicts

Bing Predicts is a prediction engine developed by Microsoft that uses machine learning from data on trending social media topics (and sentiment towards those topics), along with trending searches on Bing. It predicts the outcomes of political elections, popular reality shows, and major sporting events. Predictions can be accessed through the Bing search engine.[1]

History

The idea for a prediction engine was first suggested by Walter Sun, Development Manager for the Core Ranking team at Bing, when he noticed that school districts were more frequently searched before a major weather event in the area was forecasted, because searchers wanted to find out if a closing or delay was caused. He concluded that the time and location of major weather events could accurately be predicted without referring to a weather forecast by observing major increases in search frequency of school districts in the area. This inspired Bing to use its search data to infer outcomes of certain events, such as winners of reality shows.[1] Bing Predicts launched on April 21, 2014. The first reality shows to be featured on Bing Predicts were The Voice, American Idol, and Dancing with the Stars.[2]

The prediction accuracy for Bing Predicts is 80% for American Idol, and 85% for The Voice. Bing Predicts also predicts the outcomes of major political elections in the United States. Bing Predicts had 97% accuracy for the 2014 United States Senate elections, 96% accuracy for the 2014 United States House of Representatives elections, and an 89% accuracy for the 2014 United States gubernatorial elections. Bing Predicts is also making predictions for the results of the 2016 United States presidential primaries.[3] It has also done predictions in sports, including a perfect 15 for 15 in the 2014 World Cup,[4] leading to positive press such as a Business Insider story on its successes[5] and a PC World article on how Microsoft CEO Satya Nadella did well in his March Madness bracket entry.[6]

References

This article is issued from Wikipedia. The text is licensed under Creative Commons - Attribution - Sharealike. Additional terms may apply for the media files.