Teach computer traffic forecast like weather

Researchers are looking to use traffic videos in Berlin, Istanbul and Moscow along with big-data technology to simulate the brain to forecast traffic situations.

According to the Washington Post, the artificial intelligence (AI) research institute, founded by Here Technologies, a German company providing location and transportation data, organized a traffic forecast contest in December with the results. obtained with amazing accuracy.

Picture 1 of Teach computer traffic forecast like weather
A busy route during Christmas and New Year holidays in Moscow, Russia.(Photo: AP).

While the driver's surprise decision can easily interfere with traffic forecasts, the results of the competition show that sophisticated calculation settings known as 'neural networks' can Be aware of patterns that are difficult to detect from huge data stores.

And this could be significant in judging how vehicles travel from place to place, as well as how researchers analyze the interaction between traffic and environmental concerns. .

Here, Michael Kopp, Here's lead researcher and co-director of the Advanced Institute for Advanced Intelligence in Austria, stressed that "there is no theory to build a successful neural network" , when referring to A machine-learning model that mimics the way researchers once thought of how the brain works, computers are provided with a large amount of data, and "learn" to find certain models.

He said the researchers collected vast amounts of traffic data from the major cities of Germany, Turkey and Russia.

The amount of data for many months has been color-coded with a speed index in green, the direction of travel in blue and traffic in red. Research teams have started designing software tools to collect models from long videos. They are evaluated based on the ability to predict the traffic situation after data processing provided by the means directly, specifically after 5, 10 and 15 minutes.

The organizers said the top groups were getting amazingly accurate results. The wrong rate of the winner, Korean researcher Sungbin Choi, was surprisingly small, less than 1%.