Big data competition maps car reliability
Kaggle and its network of 25,000 PhD students analyse what second hand cars are most likely to reach their new owners with defects.
A competition to find out, using big data, which second hand cars will be the most reliable has come to a close – the answer is orange coloured ones.
The competition, which ran from September last year to early January this year, was the latest to be organised by the US company Kaggle. Over 800 individuals educated to PhD level or equivalent across more than 500 teams submitted entries for the competition. They were given what Kaggle described as a dozen large car boots’ worth of data to work with in order to find the answer.
The winning team, taking home the top prize of $5,000 (£3,200), was Marcin Pionnier of Sollers Consulting and Xavier Conort of Gear Analytics. Through big data analytics, they found that, although rare, orange cars were the most likely to reach subsequent owners defect-free.
The worst cars to buy second hand, they found, are convertibles and those with modifications such as expensive stereo systems or alloy wheels. The worst colours to buy are red and purple, according to the team’s findings.
Kaggle, which has a network of over 25,000 PhD students, has previously run a competition for NASA, the Royal Astronomical Society, and the European Space Agency to find dark matter in the universe. That competition was won by Martin O'Leary, a PhD student from the University of Cambridge.
Other companies have sponsored competitions in order to develop new working models.
While a competition is live, Kaggle publishes a real-time leader board of progress, motivating competitors to continue to further their entries along the way.
Other leading competitions ongoing at the moment include developing a method to work out which people will be admitted to hospital within the next year, carrying a $3m (£1.9m) prize fund, and one to develop an automated scoring system for student essays, backed by $100,000 (£63,500).
