Parallelized extraction of traffic state estimation rules based on bootstrapping rough set

As an important application in advanced traveler information system (ATIS), traffic state estimation can be implemented by rough set theory to extract traffic rules from original transportation data and converted traffic feature data. The real-time collected transportation data, which can respond to the exceptional cases timely, are used to update the rules bootstrappingly. Since the whole process centers its computation intensity on computing the attribute significance of rough set, we adopt message passing interface (MPI) to parallelize that computation-intensive part to improve the efficiency. The experiments on accuracy of feature conversion and efficiency of rule extraction show that our implementation can achieve high accuracy in comparison with direct conversion and historical conversion, and increasingly high efficiency when executing on more cluster nodes.