Dimensionality Reduction and Machine Learning-Based Crash Severity Prediction Using Surrogate Safety Measures
摘要
This paper presents a mechanism for using Machine Learning (ML) methods to extract the information efficiently from the input data collected in the field for predicting crash severity predictions. The output from the EVT engine subsequently can be used for approximately predicting traffic crashes. In our study, we use Principal Component Analysis (PCA) and Multidimensional Scaling techniques to obtain reduced dimensionality of the information obtained from multiple variables. Logisitic and Poisson regressions are used on the reduced number of variables which are the principal components, for crash severity predictions.