Classifier prediction evaluation in modeling road traffic accident data

This paper illustrates the research work in exploring the application of data mining techniques to aid in the prediction of road accident patterns related to pedestrian characteristics. It also provides insight into pedestrian accidents by uncovering their patterns and their recurrent underlying characteristics in order to design defensive measures and to allocate resources for identified problems. In this study the Decision Tree algorithms viz. Random Tree, C4.5, J48 and Decision Stump are applied to a database of fatal accidents occurred during the year 2010 in Great Britain. We also used K-folds Cross-Validation methods to measure the unbiased estimate of the four prediction models for performance comparison purposes.