MPI-Based Twi-extraction of Traffic State Evaluation Rules

Abstract

Through converting transportation data into some conditional attributes and one decision attribute which constitute the decision table, we use Rough Set theory (RS) to extract rules for traffic state evaluation. This method, named twi-extraction, combines the first extraction by the confidence threshold and the second extraction on the eliminated rules by the matching accuracy. Since the computational intensity is mainly placed onto the attribute significance computation of twi-extraction, Message Passing Interface (MPI) is adopted to parallelize it for acceleration. The experimental results show that by comparing the twi-extraction with the first extraction and pseudo twi-extraction, our MPI-based implementation can achieve both higher matching accuracy and higher computing efficiency.