Date of Graduation
Master of Science in Industrial Engineering (MSIE)
Justin R. Chimka
Second Committee Member
Social sciences, Applied sciences
The trucking industry and truck drivers play a key role in the United States commercial transportation sector. Accidents involving large trucks is one such big event that can cause huge problems to the driver, company, customer and other road users causing property damage and loss of life. The objective of this research is to concentrate on an individual transportation company and use their historical data to build models based on statistical and machine learning methods to predict accidents. The focus is to build models that has high accuracy and correctly predicts an accident. Logistic regression and penalized logistic regression models were tested initially to obtain some interpretation between the predictor variables and the response variable. Random forest, gradient boosting machine (GBM) and deep learning methods are explored to deal with high non-linear and complex data.
The cost of fatal and non-fatal accidents is also discussed to weight the difference between training a driver and encountering an accident. Since accidents are very rare events, the model accuracy should be balanced between predicting non-accidents (specificity) and predicting accidents (sensitivity). This framework can be a base line for transportation companies to emphasis the benefits of prediction to have safer and more productive drivers.
Francis Xavier, E. (2016). Safety Performance Prediction of Large-Truck Drivers in the Transportation Industry. Graduate Theses and Dissertations Retrieved from https://scholarworks.uark.edu/etd/1844
Industrial Engineering Commons, Operational Research Commons, Transportation Engineering Commons