Pengembangan Model Prediksi Kecelakaan Lalu Lintas Pada Jalan Tol Purbaleunyi

. Road traffic safety condition in Indonesia still needs improvement that road traffic accidents occur frequently and cause many people died. One way to improve road safety is by developing accident prediction model. The model relates accident frequencies with traffic flow and various roadway geometric/environment characteristics contributing to accident occurrences. The model is useful to estimate and predict safety of a road entity, identify hazardous road locations, and also to evaluate treatment effectiveness applied on hazardous road locations. This paper presents the accident prediction model for Purwakarta-Bandung-Cileunyi (Purbaleunyi) toll road. The model was developed by comparing statistical performances of Poisson, Negatif Binomial (NB), Zero- Inflated Poisson (ZIP), dan Zero-Inflated Negative Binomial (ZINB) regression models. The modeling result indicates that the model which is developed using the NB regression model is the best. Based on the developed model, the accident frequency has positivie correlations with annual average daily traffic, degree of curvature, and presence of median with height less than 1.75 m and width less than 2.5 m. The developed model was subsequently applied to identify and rank hazardous road segments. The result shows that KM 92-93 (direction to Cileunyi) is the most hazardous segment at Purbaleunyi toll road