University in Tokyo releases Japan ‘Bear encounter AI prediction map’ online

Sign up now: Get STs newsletters delivered to your inbox

The map forecasts bear-related risks using regional environmental data and reported sightings.

PHOTO: REUTERS

TOKYO - A research team at Sophia University in Tokyo has released an online “Bear Encounter AI Prediction Map” that estimates the likelihood of human-bear contact in 19 regions, aiming to reduce the frequency of attacks.

The map uses artificial intelligence to forecast bear-related risks in areas such as Sapporo in Japan’s northernmost Hokkaido and Akita Prefecture in Honshu, the country’s main island, drawing on regional environmental data and reported sightings.

The team hopes the tool will help residents, tourists and local governments better prepare for bear activity in their areas, as

attacks have become a growing concern

Associate Professor Yusuke Fukazawa, a data science expert, and his team developed an AI-based tool that predicts bear-human contact risks by analysing one to three years of sighting data, forest conditions, road layouts and population density.

The analysis showed higher risks along the foothills of mountainous areas, river corridors where bears can move unseen, and roads in remote valleys, while communities with ageing populations also tended to face an elevated chance of bear encounters.

Based on the system, the team released on Oct 20 a prediction map that colour-coded one-square-kilometer areas into five levels, from low to very high, representing 0 - 20 per cent to 80 - 100 per cent estimated probability of bears being in the areas.

In areas where bears have actually been seen, the model found accuracy levels varied depending on regional conditions across the surveyed locations.

Assoc Prof Fukazawa expressed eagerness to improve the