TMA (track mounted attenuators) involved work zones safety analysis and modeling, using machine learning to predict crash severity and crash frequency in the state of Missouri
When talking about road maintenance safety, it is an unavoidable talking about the TMA usage, as it located in the work zone intended to reduce the damage while crash happened and at the same time protect the related construction. Many scholars focused on derived use of safety performance functions for SWZs safety level judgement and for MWZs all existing researches are still stays at focusing working TMAs related factors. This research use TMA-related crash data in Missouri from 2011 to 2016, and through those recorded crash reports, combine using MWZs working schedules, figure out which factors and under which situations are common exist through all recorded crash both in MWZs and SWZs. Differential analysis model was explored and built in this research for a detailed knowing and referring the real reasons behind existing data. (Abstract in topic1) Machine learning is been widely used in all walks of life. Unlikely the traditional mathematic models and regression models by using both math and statistical knowledge, Machine learning performs high accurate results based on a mimic of human's brain and large data experience analysis thinking by computer. Using machine learning model for crash severity prediction and crash frequency prediction is a new thought for majority existing machine learning models using for crash prediction are supervised model or with low accuracy. This paper will use unsupervised model -- LSTM to achieve "global usage" through input whole Missouri data generated by rules and shows how sever the crash will occur under specific conditions, as well as predict crash frequency in coming years under different environment conditions. (Abstract in topic2)
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 License.