Title

Session 13: Predicting Automobile Accident Severity and Hotspots Using Multinomial Logistic regression

Presenter Information/ Coauthors Information

Zhuoyu Yang, Minot State University

Presentation Type

Oral

Student

Yes

Abstract

Americans are now driving more than ever [1]. In 2010, close to 33,000 lives were lost and another estimated 3.9 million people were injured in automobile accidents; all things considered, these accidents accounted for $836 billion in damages [2]. Since then, the rate of automobile-related deaths per 100 million miles traveled has not shown signs of improvement [3]. This research expands upon a previous year’s poster presented at the South Dakota State University Data Science Symposium 2019 [4]. While the previous research focuses on a data visualization of automobile accident hotspots on a map based on the severity and frequency of accidents, this research aims to train a multinomial logistic regression machine learning model using data related to weather conditions, speed limit, and GPS coordinates to predict the severity of automobile accidents. The development of such a machine learning model can help inform emergency services better manage resources in anticipation of potential automobile accidents based on prevailing weather conditions, speed limit along a stretch of road, and location data. An updated version of the previous dataset will be used. This dataset contains approximately 1.5 million automobile accident data points, collected over a span of over four years, from February 2016 to December 2020.

Start Date

2-8-2022 3:30 PM

End Date

2-8-2022 4:25 PM

This document is currently not available here.

Share

COinS
 
Feb 8th, 3:30 PM Feb 8th, 4:25 PM

Session 13: Predicting Automobile Accident Severity and Hotspots Using Multinomial Logistic regression

Dakota Room 250 A/C

Americans are now driving more than ever [1]. In 2010, close to 33,000 lives were lost and another estimated 3.9 million people were injured in automobile accidents; all things considered, these accidents accounted for $836 billion in damages [2]. Since then, the rate of automobile-related deaths per 100 million miles traveled has not shown signs of improvement [3]. This research expands upon a previous year’s poster presented at the South Dakota State University Data Science Symposium 2019 [4]. While the previous research focuses on a data visualization of automobile accident hotspots on a map based on the severity and frequency of accidents, this research aims to train a multinomial logistic regression machine learning model using data related to weather conditions, speed limit, and GPS coordinates to predict the severity of automobile accidents. The development of such a machine learning model can help inform emergency services better manage resources in anticipation of potential automobile accidents based on prevailing weather conditions, speed limit along a stretch of road, and location data. An updated version of the previous dataset will be used. This dataset contains approximately 1.5 million automobile accident data points, collected over a span of over four years, from February 2016 to December 2020.