Road Accident Severity Classification

Introduction

This study aims to propose an approach for road traffic severity classification using machine learning techniques. The objective is to develop a fast and efficient system for identifying the severity of road accidents and improving emergency response. We use a dataset of road accidents and their severity levels to train the model and classify the severity of future accidents.

System Requirements

We use the following system requirements for this project:

• Python 3.x
• Ram >= 4 GB
• Storage >= 20 GB
• OS: Windows 11

Tools and Technologies

We use the following tools and technologies for this project:

• Python 3
• Jupyter Notebook
• Pandas
• Numpy
• Matplotlib
• Seaborn
• Scikit-learn

activate and create conda env (recommended)

• conda create -n devenv python=3.10
• conda activate devenv
• conda install numpy matplotlib pandas scikit-learn seaborn jupyter imblearn

Dataset

The dataset used for this project is the Road Accidents Severity Dataset dataset from Kaggle. It contains 32 features and 12k records of road accidents.

Data Preprocessing

We perform the following preprocessing steps on the dataset:

• Remove unnecessary features
• Remove records with missing values
• Convert categorical features to numerical features
• Normalize the dataset

Exploratory Data Analysis

We perform the following exploratory data analysis on the dataset:

• Plot the distribution of the target variable
• Plot the distribution of the features
• Plot the correlation matrix of the features

Model Training

We train the following machine learning models on the dataset by:

• K - Nearest Neighbours (KNN)
• Decision Tree Classifer
• Random Forest Classifier
• Support Vector Classifier (SVC)
• Logisitic Regression

Model Evaluation

We evaluate the performance of the models using the following metrics:

• Accuracy
• F1 Score
• Confusion Matrix

Comparative Analysis

Dashboard using Power BI

Conclusion

In the context of road accident severity classification it has been observed that number of vehicles involved and number of casualties consistently emerge as significant factors indicating their strong influence on predicting the severity of accidents.

By incorporating the number of vehicles involved and the number of casualties into the predictive models machine learning algorithms can enhance road accident severity classification systems. This can play a important role in improving road safety measures, developing targeted interventions, and implementing proactive strategies to prevent accidents and minimize their consequences.