This research analyzes the frequency of motorcycle traffic accidents occurring on national roads in Banda Aceh City. Its primary aim is to determine the variables that contribute to motorcycle accidents and to establish the most appropriate predictive model for explaining how these variables relate to accident occurrences. Accident data from 2018–2020 were collected from 14 road segments and classified into 27 observation points, complemented by traffic flow and geometric road characteristics. The analysis included multicollinearity testing (VIF < 10), Poisson regression modeling, overdispersion testing, negative binomial modeling, and model selection using the Akaike Information Criterion (AIC). The Poisson model exhibited overdispersion (Deviance/df = 5.86), leading to the use of the negative binomial model, which reduced the ratio to 1.16 and yielded a lower AIC value (191.49) compared to the Poisson model (253.73). Results indicated three significant variables at α = 0.05: traffic volume, lane width, and the presence of U-turns. A 1% increase in traffic volume raised accident expectations by 1.0694 times, while a 1-meter lane width increase reduced accident expectations to 0.413 times; conversely, U-turn presence increased accidents by 2.217 times. These findings highlight the importance of improving road capacity through lane widening, regulating U-turn locations, and managing driving behavior on high-volume roads to reduce accident risk.