Zero-Inflated Negative Binomial model to Overcome Excess Zeros Count in Motorcycles Road Accident

Motorcycle is becoming one of the most important transportation modes and its intensity usage is increasing tremendously on the roadway. Therefore, problems of its reliability and safety are highly well-defined and discussed. In this paper, the most used model of count data for accident modeling namely Poisson and negative binomial regression are presented along with the zero-augmented model namely zero-inflated Poisson, hurdle Poisson, zero-inflated negative binomial and hurdle negative binomial will be fitted to a real motorcycle road accident data. The model validation result shows that zero-inflated negative binomial fit the data well and the highest traffic offenses and locations factors are determined.