Modeling traffic accidents in Saudi Arabia using classification techniques

Private cars are the main form of transportation in Saudi Arabia. Although the country is one of the high-income countries in the world, it has an increasing rate of road accidents. Official statistics indicate that the capital city, Riyadh, has the highest percentage of road accidents in the country. We believe that a better understanding of road accident records can help the authorities take effective measures. In this paper, we apply data mining in order to understand the factors leading to car accidents severity in Riyadh. Three classification techniques are used: CHAID, J48, and Naive Bayes. The performance of all the obtained models is evaluated and compared. The resulting models highlight the danger of distraction while driving. The age of the car is also found to be an important factor. Accidents with older cars are more likely to result in injuries or deaths. In terms of accuracy, all models have comparable accuracy values. However, poor performance is observed in predictions related to Death class. Which suggests the causes of death in road accidents are complicated and requires more investigation.