# Analyzing NYC traffic data using Pandas

I wrote a post earlier about how to analyze data using Pandas. In that post, I just introduced some simple Pandas functions to analyze some random data. In this post, I am using real world traffic data from NYC. NYC has made a lot of data available to public recently on this website.  There is plenty of quality data for you to play around with. I chose to look at the traffic data which you can download from here.

I performed the analysis using iPython Notebook which I have embedded here.

In [226]:
import pandas as pd
import numpy as np
pd.set_option('display.max_rows', 12)

In [227]:
# Load a csv directly into a dataframe
# parse_dates parameter converts the DATE and TIME columns into DATETIME format, resulting in a combined column called DATE_TIME
df = pd.read_csv("c:\NYPD_Motor_Vehicle_Collisions.csv",parse_dates=[['DATE', 'TIME']])

# Remove spaces from column names
cols = df.columns
cols = cols.map(lambda x: x.replace(' ', '_') if isinstance(x, (str, unicode)) else x)
df.columns = cols

In [228]:
# Delete the columns we don't care about
df = df.drop(['LATITUDE', 'LONGITUDE', 'LOCATION', 'UNIQUE_KEY','ON_STREET_NAME','CROSS_STREET_NAME','OFF_STREET_NAME'], axis=1)

In [229]:
# Lets do some aggregation
# Find total by year
df.groupby(df.DATE_TIME.dt.year).sum()

Out[229]:
ZIP_CODE NUMBER_OF_PERSONS_INJURED NUMBER_OF_PERSONS_KILLED NUMBER_OF_PEDESTRIANS_INJURED NUMBER_OF_PEDESTRIANS_KILLED NUMBER_OF_CYCLIST_INJURED NUMBER_OF_CYCLIST_KILLED NUMBER_OF_MOTORIST_INJURED NUMBER_OF_MOTORIST_KILLED
2012 836355612 27445 137 5907 72 2210 6 19328 59
2013 1682824973 55112 297 11984 176 4074 11 39054 110
2014 1688700297 51198 262 11027 133 3997 20 36171 109
2015 759746044 20931 103 4284 53 1319 4 15328 46
In [230]:
# Now we will look at data for a specific year
# Find total number of persons injured and killed by month in 2014
df[df.DATE_TIME.dt.year == 2014].groupby(df.DATE_TIME.dt.month).sum()[['NUMBER_OF_PERSONS_INJURED','NUMBER_OF_PERSONS_KILLED']]

Out[230]:
NUMBER_OF_PERSONS_INJURED NUMBER_OF_PERSONS_KILLED
1 3920 23
2 3106 13
3 3853 18
4 4048 22
5 4658 26
6 4985 22
7 4603 29
8 4476 22
9 4464 28
10 4718 20
11 4242 26
12 4125 13
In [231]:
# Sort the data in descending order on NUMBER_OF_PERSONS_KILLED by zip code
df.groupby(df.ZIP_CODE).sum().sort(['NUMBER_OF_PERSONS_KILLED'],ascending=0)

# Looks like Zip Code 11236 is the most dangerous

Out[231]:
ZIP_CODE NUMBER_OF_PERSONS_INJURED NUMBER_OF_PERSONS_KILLED NUMBER_OF_PEDESTRIANS_INJURED NUMBER_OF_PEDESTRIANS_KILLED NUMBER_OF_CYCLIST_INJURED NUMBER_OF_CYCLIST_KILLED NUMBER_OF_MOTORIST_INJURED NUMBER_OF_MOTORIST_KILLED
ZIP_CODE
11236 56123820 1945 18 304 7 67 0 1574 11
11434 52722174 1448 12 180 7 43 0 1225 5
11203 61818154 2311 11 404 4 123 0 1784 7
10025 30445925 602 10 227 9 91 1 284 0
11206 49127104 1432 10 363 6 255 1 814 3
10022 66987048 834 9 351 9 125 0 358 0
... ... ... ... ... ... ... ... ... ...
10307 4225870 90 0 11 0 0 0 79 0
10282 400998 6 0 6 0 0 0 0 0
10281 236463 3 0 1 0 2 0 0 0
10280 894360 14 0 3 0 6 0 5 0
10111 111221 0 0 0 0 0 0 0 0
11697 713517 19 0 1 0 6 0 12 0

190 rows × 9 columns

In [232]:
# Graph number of pedestrians/motorist killed by year
df.groupby(df.DATE_TIME.dt.year).sum()[['NUMBER_OF_PEDESTRIANS_KILLED','NUMBER_OF_MOTORIST_KILLED']].plot(kind='bar')

# 2013 was the worst year for pedestrian deaths.

Out[232]:
<matplotlib.axes._subplots.AxesSubplot at 0x9fa47160>
In [233]:
# Lets look at year 2014 and see how the deaths varied by Borough
df.groupby(df.BOROUGH).sum()[['NUMBER_OF_PEDESTRIANS_KILLED','NUMBER_OF_MOTORIST_KILLED']].plot(kind='bar')

# As we can see, most number of pedestrians died in Brooklyn than any other.
# Staten Island had the least number of deaths for both pedestrians and motorists. Surprised?

Out[233]:
<matplotlib.axes._subplots.AxesSubplot at 0x94532940>
In [240]:
# How did the deaths of pedestrians vary by year in Manhattan?
df[df.BOROUGH == 'MANHATTAN'].groupby(df.DATE_TIME.dt.year).sum()[['NUMBER_OF_PEDESTRIANS_KILLED']].plot(kind='bar')

Out[240]:
<matplotlib.axes._subplots.AxesSubplot at 0x941e0668>
In [237]:
# Distribution of deaths by borough which have VEHICLE_TYPE_CODE_1 as passenger vehicle
df[df.VEHICLE_TYPE_CODE_1 == 'PASSENGER VEHICLE'].groupby(df.BOROUGH).sum()[['NUMBER_OF_PERSONS_KILLED']].plot(kind='bar')

Out[237]:
<matplotlib.axes._subplots.AxesSubplot at 0x86b6f2b0>

I will try to do a similar analysis with q at some time in future (minus the graphs, of course).