Click here to download detailed Report

Problem Statement

The primary goal of the project was to analyze traffic accident data in the US from 2016 to 2023 using the ”US Accident (2016-2023)” dataset and find out relevant features that can be used to build a predictive system in the future.

• Analyse the primary causes of accidents and their associated risk levels
• Identify key factors that most reliably indicate the occurance of a major accident
• Find out relevant features that could help forecast both the probability and severity of accidents before they happen.
• Share the insights with stakeholders about where accidents commonly happen, patterns in casualties, and how weather and environmental conditions influence accident occurrence.

Approach

• Dataset Overview
• Import relevant libraries
• Load the dataset
• Data Preprocessing
  • Removing Unnecessary Features
  • Transforming Datetime Feature
  • Handling Missing Values
  • Outlier Treatment
• Exploratory Data Analysis & Feature Engineering
  • Resampling
  • Time-Based Feature Analysis
  • Location-Based Feature Analysis
  • Weather Feature Analysis
  • Point of Interest Features Analysis
• Feature Selection
• Further Analysis
• Key Insights and Recommendations

Exploratory Data Analysis & Feature Engineering

Time-Based Feature Analysis

• When do most severe accidents occur during the year?
• How does accident frequency vary between weekdays and weekends?
• What time of day sees the most accidents?
• Are night accidents different from day accidents?

Location-Based Feature Analysis

• Which states have the highest accident counts?s
• Which states have the most severe accidents?
• Does public transit usage in a county affect accident rates?
• Which types of roads are most dangerous?
• Are accidents concentrated in specific cities?
• Which timezone sees the most accidents?

Weather Feature Analysis

• Which weather conditions are most dangerous for driving?
• What role do visibility, pressure, and wind speed play?

Point of Interest (POI) Feature Analysis

• How do traffic signals and crossing impact accident severity?
• Are there any specific infrastructural impact on accident severity?

Key Takeaways

• Summer months + weekends + nighttime = Highest risk for severe accidents
• Georgia & Florida lead in severe accidents despite CA/TX/FL having most total accidents
• Less public transit = More severe accidents in a county
• Traffic signals & crossings cut accident severity significantly
• Interstate highways are hotspots for severe accidents
• Accidents spread across cities rather than concentrated in specific areas
• Night accidents are fewer but deadlier
• Rain/snow conditions notably increase accident severity

References

• Moosavi, Sobhan, Mohammad Hossein Samavatian, Srinivasan Parthasarathy, Radu Teodorescu, and Rajiv Ramnath. “Accident Risk Prediction based on Heterogeneous Sparse Data: New Dataset and Insights.”

• Triveni Sangama Saraswathi Edla. "Exploratory Data Analysis (EDA) of US Accidents and Prediction of Accident Severity in San Fransisco bay area"

• Medium: Ronghui Zhou, Ph.D.""How You Can Avoid Car Accident in 2020"