Tunisian policy for road safety is neither clear nor reliable. Each road organism acts on own its side without any coordination with other stakeholders and therefore without achieving results. This paper aims to facilitate the decision-making process on road safety in Tunisia through the time series analysis of road accidents. The analysis work will allow identifying better the respective weight of the factors associated with the road accident frequency. Methodologically, ARIMA (Auto-Regressive Integrated Moving Average) model is used to meet our above goal. Moreover, the methodology of Box Jenkins intervenes as a statistical solution to solve the problem of time series analysis. The results show that the time series of accidents are mainly characterized by two different periods in terms of trend. A low decrease in the number of accidents before the revolution (2011) (between 2007 and the end of 2010) and the irregular evolution in the rest of the series. Then, models are developed in accident cases and ARIMA (0, 1, 2) is identified as the best model. A three-year forecast is made using the best model and it shows that the number of road accidents would decrease due to several factors in Tunisia. Consequently, this study shows that the temporal analysis of the time series of road accidents can attract the attention of decision-makers to the importance of the application of key road safety measures in the short, medium and long term as well as the nature of the relationship between the different decision-makers horizons.