NYC Motor Vehicle Collision Data Analysis

Posted on Jun 11, 2021

Github | Shiny App

Introduction

People from all over the world come to New York City (NYC) in search of opportunities and make it their home. It is thus, one of the most populated cities in United States (US). With over 27,000 people per square mile, it has the highest population density of any major city in the US. Additionally, it leads the list of cities with best public transportation system.

And yet, it is also one of the top accident-prone cities in the country. In the year 2014, the New York Police Department (NYPD) had started a citywide traffic safety initiative called “Vision Zero”. The goal of “Vision Zero” is to eliminate traffic fatalities by collecting comprehensive data on traffic accidents. Inspired by this, I decided to analyze the motor vehicle collision data to see if we could gain some insights, that would help NYPD to take adequate measures in ensuring public safety.  

To this end, I obtained the dataset from NYC's OpenData website. The data comes from police reports, which NYPD is required to fill out when someone is killed, injured or there is a damage of at least $1000 in an accident. It is a large dataset of approximately 396 MB and reports accidents from July 2012 through May 2021. It contains about 1.7 million observations across 29 attributes, which include crash date, crash time, borough, intersection streets, latitude, longitude, number of people killed, number of people injured, number of pedestrians/ cyclists/ motorists killed and injured, vehicles involved in accidents and contributing factors. The dataset has various problems pertaining to missing values, duplicate mappings etc. and I performed several preprocessing steps before starting the exploratory data analysis. Cleaning process reduced the size of my dataset by about 30% and I also truncated the data for years 2012 and 2021 since they did not cover full years. As an end goal, I have developed a stand-alone Shiny app, which demonstrates and explains all the analysis.

Data Analysis

First, I wanted to see how the 5 boroughs of NYC differed in number of collisions. So, I plotted the total number of collisions in each borough for different timeframes. Looking at the yearly trends, I observed that Brooklyn has the highest number of crashes consistently, followed by Queens, Manhattan, Bronx, and Staten Island, respectively. In 2019, the number of accidents dropped compared to previous year in all 5 boroughs, with Bronx reporting the smallest decline of 4.3% and Staten Island reporting a whopping drop of 42.0%. This could partly be attributed to reduction in speed limit on NYC streets, which was implemented in 2018. In 2020, all the boroughs reported a steep decline in collisions, which could plausibly be accounted to the COVID lockdown.

Then, I looked at the seasonality by counting the number of collisions for each month of 8 years.  I noticed a sharp dip in accidents in all boroughs in the months of February and April. Decline in February could be attributed to fewer vehicles on road due to snow. However, the dip in April puzzled me until I found out that the police enforcement slightly increases during the early part of spring and a lot more speeding tickets are issued in April compared to preceding or following months. And this explains the drop.

Does day of the week impact the number of accidents? The plot revealed Friday to be the worst day for driving as the number of accidents reported shot up compared to other days of the week in all 5 boroughs while weekends appear relatively safe.

Further, I went onto look at the hourly trends, and noticed a sharp dip at around 3:00 pm in all boroughs. At a higher resolution, i.e., in a half-hourly plot, this dip becomes prominent at around 3:30 pm illustrating a decline in the number of accidents around that time. This was confounding as I pondered over the reasons. Upon some research, I found out that the police shift changes around 3:40 pm. I suspect that during this hour, officers from the previous shift are winding down and newer shift are warming up, so neither is keeping a good record of the accidents, due to which it appears that cases are less but chances are that it is the reporting that is less. Likely all the cases around that time get reported at or after 4 o'clock.

Next, I wanted to see who the victims are in these accidents, how many of them are killed and how many injured. So, I plotted percentage of people killed/injured in each borough each year. Overall, Staten Island seems to have had greatest percentage of people killed out of the total number of accidents per year in that borough. Manhattan, on the other hand, has the lowest percentage of accidents that prove fatal. This might be the result of congestion and frequent stops in Manhattan, which reduce vehicle speed. In addition, the presence of some of the best hospitals in Manhattan likely makes a difference. Interestingly, the number of fatalities went up in 2020 in all boroughs although the number of collisions were fewer. In my opinion, this may have resulted from healthcare resources getting directed to COVID patients.

Amongst the victims, Manhattan shows the highest percentage of pedestrians dying in collision while cyclists appear safest in all boroughs.

Of the percentage of people injured, Bronx and Brooklyn are in the lead. Manhattan is again at the bottom. Further, it is to be noted that percent injured shot up in 2019 in Staten Island. However, in absolute numbers, there was no perceptible increase in the number of people injured. Since, the number of collisions had dropped almost 40% in 2019 in Staten Island, the base was much lower which led to a sharp climb in percentage. Furthermore, as  with fatalities, percentage injured also increased in 2020.

Subsequently, I wanted to examine the factors that have contributed to all the accidents. A bar plot of causal factors against percent accidents revealed that in majority of cases (~26 -38%), the cause is "Unspecified" in all boroughs.  This should be addressed by keeping a better record of the exact cause of accident in order to eliminate the causal factors. In case, NYPD is not able to record a cause until it is precisely determined by insurance companies, a system of communication needs to be established between the insurance companies and NYPD. The next big cause in all boroughs is "Driver inattention/ Distraction" (~18-27%), where Manhattan is at the upper end of the range. The city authorities should ensure that there are no distractions like billboards on the streets. If people's fears, anxieties and stress-levels could be addressed with meditative driving, it might restore their focus and attention on road.

Finally, I have highlighted the intersections that have witnessed at least one accident per week on average. Red dots on the map below show these areas and in the Shiny app, one can click on them to see the street intersections with count of accidents that have occurred there in the last 8 years. For example, as shown below, intersection of the West Fordham Road and Major Deegan Expressway in Bronx has witnessed a total of 675 accidents from 2013 through 2020.  I have identified 22 such intersections across NYC. 

NYC Map showing the hotspot areas of accidents

Conclusion

With this information in hand, the city authorities and NYPD should start implementing changes in at least these extremely risky areas in NYC. Additionally, there should be better reporting of the cause of accident so measures can be taken accordingly. A process should be set up to obtain a detailed report on the cause of accident from insurance companies. Finally, decisions that helped bring the counts of collisions down in 2019 should be further pursued. 

About Author

Ireena Bagai

Ireena is a current NYC Data Science Academy fellow with several years of research experience. Prior to this, she worked as a Science Advisor/Patent Agent at a law firm in NYC where she analyzed patents and advised on...
View all posts by Ireena Bagai >

Leave a Comment

No comments found.

View Posts by Categories


Our Recent Popular Posts


View Posts by Tags

#python #trainwithnycdsa 2019 airbnb Alex Baransky alumni Alumni Interview Alumni Reviews Alumni Spotlight alumni story Alumnus API Application artist aws beautiful soup Best Bootcamp Best Data Science 2019 Best Data Science Bootcamp Best Data Science Bootcamp 2020 Best Ranked Big Data Book Launch Book-Signing bootcamp Bootcamp Alumni Bootcamp Prep Bundles California Cancer Research capstone Career Career Day citibike clustering Coding Course Demo Course Report D3.js data Data Analyst data science Data Science Academy Data Science Bootcamp Data science jobs Data Science Reviews Data Scientist Data Scientist Jobs data visualization Deep Learning Demo Day Discount dplyr employer networking feature engineering Finance Financial Data Science Flask gbm Get Hired ggplot2 googleVis Hadoop higgs boson Hiring hiring partner events Hiring Partners Industry Experts Instructor Blog Instructor Interview Job Job Placement Jobs Jon Krohn JP Morgan Chase Kaggle Kickstarter lasso regression Lead Data Scienctist Lead Data Scientist leaflet linear regression Logistic Regression machine learning Maps matplotlib Medical Research Meet the team meetup Networking neural network Neural networks New Courses nlp NYC NYC Data Science nyc data science academy NYC Open Data NYCDSA NYCDSA Alumni Online Online Bootcamp Online Training Open Data painter pandas Part-time Portfolio Development prediction Prework Programming PwC python Python Data Analysis python machine learning python scrapy python web scraping python webscraping Python Workshop R R Data Analysis R language R Programming R Shiny r studio R Visualization R Workshop R-bloggers random forest Ranking recommendation recommendation system regression Remote remote data science bootcamp Scrapy scrapy visualization seaborn Selenium sentiment analysis Shiny Shiny Dashboard Spark Special Special Summer Sports statistics streaming Student Interview Student Showcase SVM Switchup Tableau team TensorFlow Testimonial tf-idf Top Data Science Bootcamp twitter visualization web scraping Weekend Course What to expect word cloud word2vec XGBoost yelp