Abstract
The underreporting scenario is claimed to be the source of the extra zeros in traffic accident data. This leads to a latter problem in which the fitted statistical model may not be able to produce correct and reliable estimates. Understanding the root of problem as to what is the main cause of the underreporting scenario is essential to assist on the decision making process in traffic accident analysis. In this study, 200 Malaysian drivers were interviewed on their sentiments towards this issue. Their opinions on the causes of underreporting scenario are investigated then assessed using text analyses. First, the Latent Dirichlet Allocation text modelling is employed to find the underlying themes in the reasons of not reporting a traffic accident. Then, the polarity of the topics is measured using a lexicon based sentiment analysis. Results showed that majority Malaysian drivers (80.5%) consider that reporting a minor or non-fatality accident is not important and can be neglected. The decision is due to the fact that of complicated and time consuming reporting process. The drivers are also asked on their opinion after the consequences of underreporting are informed to them. The polarity of their answers shifted to more positive in which 71% drivers will report an accident that occur in the future.
Highlights
Traffic accident analysis is vital to society and country as its impact can be huge physically and economically
The presence of extra zeros in accident count data is associated with the underreporting scenario [6,7]
As mentioned in the previous section, the data in this study is based from the interview session conducted on 200 Malaysian drivers
Summary
Traffic accident analysis is vital to society and country as its impact can be huge physically and economically. It is essential to understand influencing factors and predicting future outcomes in terms of traffic accident frequency and severity. To achieve these objective, statistical analyses have been conducted by utilising the reported traffic accident data as can be found in [1,2,3]. As mentioned by [45], the accident count data often exhibit extra zeros which requires the need of zeroaugmented models. The presence of extra zeros in accident count data is associated with the underreporting scenario [6,7]. We aim to explore the drivers’ opinion on the underreporting issue and what leads them to not reporting an accident
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
