Taxis in Japan Are Using Artificial Intelligence to Predict Ride Requests

NTT Docomo's predictive technology increased business for one taxi driver by 20 percent.

Long Exposure Photography
Jiangang Wang#99287—Moment Editorial/Getty Images

On a Friday evening, it's fairly easy to guess the downtown hot spots where taxis likely are needed. But most of the time, cab drivers have little more than intuition to go on to find their next fare. That could change for cab drivers in Japan when NTT Docomo commercializes its artificial intelligence technology, which predicts where ride requests will be.

Artificial Intelligence accurately predicts demand, increases business

Using ridership data from 4,425 cabs, along with other factors such as weather and mobile phone data and locations, NTT Docomo trained its AI system to predict where localized ride demand will be in 30 minutes within a 500-square-meter area. Drivers participating in its trials used tablets that provided updated ridership data every 10 minutes, which gave drivers enough time to reposition their vehicle based on anticipated requests. 

Docomo says that its forecasts were more than 80 percent accurate 90 percent of the time, according an article in Nikkei Asian Review. One cabdriver reported that his business increased by 20 percent. Docomo may try to commercialize the technology later this year. 

Uber also tracks user data

The Japanese wireless carrier isn't the only company mining data from mobile phones to make it easier to match drivers with riders. Last year, Uber began tracking its customers' GPS location for five minutes after rides ended, and redeveloped its application to enable it to mine users' calendars and contacts for relevant data. This information can provide hot maps of local demand, or help predict surges in ride requests.

Although many privacy advocates are concerned about personal data sharing, Uber users must voluntarily consent to provide this information, which ultimately may end up benefiting them and public. Depending on how it's used, location and event data could shorten Uber ride wait times and reduce the amount of time operators drive aimlessly around prior to picking up a fare—which could, in turn, lessen congestion and reduce tailpipe emissions.