This project aim is to explore Accident data and build a model to predict accident severity
Accident in traffic is an area of primary concern to society because it leads to loss of property and even loss of lives. It is therefore important to identify the key factors of accident severity, to avoid and prevent such loss and to improve safety. A model to predict accident severity is built using these predictors as input vectors. Once the model is validated using a Machine Learning algorithm approach, we can be confident in predicting accident severity. In this proactive approach, the results of the analysis would be useful to various Entities like the Police and Insurance Companies. The goal is to reduce the fatalities and economic losses from accidents.
The data used in this analysis is a collection of traffic records from the Seattle Department of Transportation (SDOT) Traffic Management Division. It consists of 194,673 observations and 38 variables. The variable SEVERITY CODE is the dependent variable to be predicted. It has 2 values: 1- Property Damage; 2- Injury. We would like to identify the variables (among the 37 left ) that increase the severity of accident fatality. Some variables are numerical, and some are categorical. Among others, the followings are potential variables of interest: Location, Weather condition, Car Speeding, Light condition, Road condition, Junction type, Number of people involved, and Number of vehicles involved in the accident.