*5.2. Additional Consideration*

Although the main factors determining the degree of offender and victim injury were identified with this study, it is necessary to find out why there exist the differences in accuracy between the two, notwithstanding the same dataset were used for the analysis. It is because the principal cause of injury to the offender is the vehicle type whose data was originated from the TAAS that requires no data correction, but the cause of injury to the victim (speed) is the data obtained from the TOPIS data that entail a lot of missing values (see Section 3.1). It is judged that the difference in the factors that determine the injury occurred in the original data (TAAS & TOPIS), lead to the differences in accuracy. However, the accuracy derived from this analysis is higher than that of existing statistics-based traffic accident prediction models (Poisson regression model, negative binomial regression model, etc. [56–59]. This is judged to be meaningful, as car accidents can be prevented in advance by predicting traffic accident injuries through DNN learning.
