Abstract
Local governments in Japan are facing significant challenges due to demographic changes, particularly population decline and an aging society. These shifts have led to workforce shortages and financial constraints, pushing municipalities to adopt artificial intelligence (AI) solutions to improve operational efficiency and public services. A major issue observed at city halls is congestion at service counters, which results in prolonged waiting time for citizens and increased workloads for staff. This study investigates the case of Takayama City Hall in Gifu Prefecture, where data from ticket dispensers at the citizen affairs counter were used to predict congestion levels. Using LightGBM, a machine learning model, we forecasted congestion two months in advance, measured on a five-point scale. To ensure these predictions were effectively communicated to the public, we developed a user-friendly congestion forecast calendar. The calendar provides hourly congestion projections for the upcoming two months and it has been available on the Takayama City Hall website since June 2023. This allows residents to plan their visits more strategically, potentially reducing waiting time and improving satisfaction with city services. The congestion forecast calendar has received positive feedback from municipal staff, and there was a notable increase in website traffic during the Obon holiday period (in August), indicating its particular utility during peak times at service counters.
