A large-scale car stranded on a national highway in Niigata Prefecture due to record heavy snowfall.


At that time, we analyzed data searched by people in the surrounding area on Yahoo, a major Internet company.


Especially conspicuous words are "live camera", "stack" and "Echigo Kotsu".


What does that mean?

The expansion of stagnation is also in the data

NHK and Yahoo! JAPAN, based on the big data aggregated by Yahoo in a way that individuals cannot be identified, compared the congestion situation in Niigata Prefecture from the 19th to the 21st of this month with respect to the situation of stuck people and traffic jams. and analyzed the most searched words.



First, we will analyze how many people were in the area around National Route 17 from Nagaoka City to Ojiya City in Niigata Prefecture during the time period when the stranded people occurred, by dividing the area into 1km squares.

At 8:00 pm on the 19th, the area near Koshi Ohashi in Ojiya City, where people were intermittently stuck, and the area near the Katada Interchange on National Route 17 in Nagaoka City were more congested than normal. became.

Then, five and a half hours later, at 1:30 am on the 20th, the red areas indicating congestion increased along the national highway.

The data also confirmed that the stranded vehicle was extending along the national highway.

what information did you look up?

Next, what kind of information did the people in this area look for?


We examined the words searched in chronological order compared to the whole country in the area around this national highway where the stuck occurred.

On the night of the 18th, before the stagnation occurred, in addition to searches related to commercial comedy programs that were being broadcast that day and NHK's Taiga drama "13 people of Kamakura-dono", "highway traffic information" and "snow forecast" were searched. "" and "closed roads" are conspicuous, and words that look up traffic conditions and future snow information before the snow gets into full swing have begun to appear.

And from the afternoon of the 19th, when the stucks began to occur, most of the searched words were related to heavy snow.



At night, in addition to “traffic jam” and “Kashiwazaki City” where people were stuck, words such as “live camera” and “what is a stack” start to stand out.

The search word "does not work" was also seen.

On the 20th, the next day, the words "live camera", "what is stack", and "Echigo Kotsu" from the local bus company stood out.

When an expert analyzes...

Associate Professor Naoya Sekiya of the University of Tokyo, who studies disaster information, analyzed the meanings of these words.



▽ “Live cameras” are said to be playing an increasing role, as many people wanted to see the snow conditions with their own eyes and make decisions.



▽ "What is a stack?" has "towa", so it can be inferred that there were many people who did not understand the meaning of "stack" that appeared in the media and wanted to find out.



▽ Regarding bus companies, it seems that "timetable" appears in other search words, and it seems that there was interest in information about the start of bus operation, which had been completely suspended.



Associate Professor Sekiya said, ``Because the range of search word analysis is wide, it includes many words searched by people other than those who are stuck.'' It is possible to analyze whether

What is the significance of search data analysis?

Kota Tsubouchi, a senior researcher at Yahoo! JAPAN Research Institute, who conducted the data analysis, said, "By combining and analyzing two types of big data, the congestion situation and search words, we were able to find information about the area around the disaster site, even as the disaster situation changed moment by moment. Based on this case study, we are also considering a method for quicker and more detailed analysis, and will use information to help prevent and mitigate disasters. I want to find a way," he said.

*About data analysis

▽Congestion analysis


Machine learning is performed based on big data of location information collected by Yahoo in a way that individuals cannot be identified, and the difference from normal crowds in each area is quantified.

We analyzed whether the number of people in the target area was higher or lower compared to normal times for each time period.



▽ Analysis of


search words Machine learning is performed based on big data of search words collected by Yahoo in a way that individuals cannot be identified, and the difference between words searched nationwide in the target area is digitized.

In the target area, we extracted and analyzed the words that were searched more frequently compared to the whole country for each time period.

This time, an area of ​​5 km from east to west and 12 km from north to south was set as the target of the analysis so as to include the national highway of Nagaoka City.