-
Road Accidents have a huge economic and societal impact costing hundred of billions of dollar every year. A study by the Department of Transportation's, National Highway Traffic Safety Administration (NHTSA), placed a price tag of over a quarter-trillion dollars in 2010. A large part of these losses is caused by a small number of serious accidents. Reducing accidents, especially these serious accidents, is an important challenge.
-
Objective of our project is to identify the key factors affecting the accident severity. Also, train a model which predicts:
- Accident Severity - a number between 1 to 4, where 1 indicates the least impact on traffic and 4 indicates a significant impact on traffic
-
We took the literature review and references from Accident Risk Prediction based on Heterogeneous Sparse Data: New Dataset and Insights and Predicting Crash Injury Severity with Machine Learning Algorithm Synergized with Clustering Technique: A Promising Protocol
-
In the initial stages of the project, we did the EDA of the above datasets and did feature engineering to come up with the most optimal dataset.
-
Created baseline models:
- Decision Tree
- Logistic Regression
-
Main Models:
- Decision Tree
- Random Forest
- SVM
- Logistic Regression
- ADA Boost
- XG Boost
- Neural Networks
-
The files in the folder
Model Training Code
are used for creating the model dumps of the above 2+7 models and the weights are saved. -
Data and Trained models can be found here : https://drive.google.com/drive/folders/1D8hd2GPEJYF-BKjkfV1-7CILVjdi-zDt?usp=sharing