This is a project to perform exploratory data analysis (EDA) on the US accidents dataset. The dataset contains 2.8 million accident records across the US, with 47 different features. The aim of this project is to analyze three columns - cities, start time, start latitude, and longitude - to gain insights into the factors contributing to accidents and their impact.
The US accidents dataset is available on Kaggle. It is a large dataset that contains information about accidents reported in the US from February 2016 to December 2021.
The following libraries and tools were used for this project:
- Matplotlib
- Seaborn
- Pandas
- opendataset
- Folium
The analysis performed on the US accidents dataset provided the following insights:
- There is data about 11628 cities.
- Miami has the highest number of accidents (106966), followed by Los Angeles (68956), Orlando (54691), Dallas (41979) and Houston (39448).
- No data for New York, even though it is the most populous city in US.
- There are over 1110 cities with exactly one accident reported.
- Less than 5% cities (494) have more than 1000 number of accidents.
- More than 95% cities (11187) have 1000 and less than 1000 number of accidents.
- 2021 has the highest number of accidents reported (1511745).
- On an average, High Percentage of accidents occur between 1 PM to 5 PM.
- Weekdays have more accident reports than weekends.
- On weekdays the trend of number of accidents has two maximas (7AM to 9AM) and (3PM to 5PM).
- On weekends between 12 PM and 6 PM, there are more accident reports.
- More number of accidents occur in winters.
- There is some missing data for the year 2016.
- In 2017, there are more number of accidents in the starting of the year.
- In 2020, data for a month is either missing or that month has 0 number of accidents reported. This maybe due to the COVID-19 quarantine.
The US accidents dataset is available on Kaggle. Special thanks to Sobhan Moosavi, Mohammad Hossein Samavatian, and Srinivasan Parthasarathy for making this dataset publicly available.