In this project, we aim to predict the severity of accidents by considering various factors, such as location, collision type, and number of vehicles involved. The motivation behind this project is to leverage the available data to develop a computer-generated model that can warn users about potential hazards and help them make informed decisions about their travel plans.

The dataset used in this project contains information about over 194,000 accidents, with 37 columns providing details like the severity of the accident, the number of people and vehicles involved, weather and road conditions, and other relevant factors. The initial step involves data cleaning and preprocessing, where we handle missing values, convert categorical variables into numerical format, and engineer new features that could improve the model’s performance.

Exploratory data analysis is then conducted to gain insights into the dataset, such as understanding the relationship between accident severity and different factors, identifying patterns and trends, and recognizing potential outliers or anomalies. Based on these insights, a logistic regression model is developed to predict the likelihood of an accident resulting in an injury collision or a property damage-only collision. 

The performance of the model is evaluated using various metrics, including Jaccard similarity score, confusion matrix, and log loss. Finally, the project aims to provide useful recommendations and insights that can help transportation authorities, policymakers, and individual travelers make more informed decisions to enhance road safety and reduce the impact of accidents.