Analysis of Motor Vehicle Thefts in Chicago

The data about motor vehicle thefts in chicago are taken from the Chicago Data Portal



In [1]:

    
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

%matplotlib inline



In [51]:

    
mvt = pd.read_csv('../../data/mvt2015_2017.csv')
mvt.head(3)









    Out[51]:







  
    
      
      Date
      Block
      Primary Type
      Description
      Location Description
      Ward
      Community Area
      Year
      Location
    
  
  
    
      0
      05/03/2016 08:00:00 PM
      100XX S SANGAMON ST
      MOTOR VEHICLE THEFT
      AUTOMOBILE
      STREET
      34.0
      73.0
      2016
      (41.711843569, -87.646607932)
    
    
      1
      05/03/2016 11:00:00 PM
      084XX S MORGAN ST
      MOTOR VEHICLE THEFT
      AUTOMOBILE
      STREET
      21.0
      71.0
      2016
      (41.740895923, -87.648617881)
    
    
      2
      05/03/2016 04:45:00 PM
      003XX W MONROE ST
      MOTOR VEHICLE THEFT
      AUTOMOBILE
      STREET
      2.0
      32.0
      2016
      (41.88063228, -87.635935494)

Exercise 0

Inspect the mvt-data (columns, colum-types, first rows...)

How many entries are there?
How many different Location Descriptions exists?



In [48]:

    
# TODO how many entries



In [49]:

    
# TODO different Location Descriptions

Exercise 1

Draw a histogram to investigate, which locations (Location Description) are the top 10 theft hot spots horizontally.

Exercise 1.1

Use Dataframe/Series.plot to plot the histogram (barchart) horizontally.

Note:
DataFrame.plot.hist() can only work with numerical x-values. If you don't want to convert the 'Location Descriptions' to a numeric value, you may want to use Series.value_counts().plot instead.



In [5]:

    
# TODO









    Out[5]:





<matplotlib.axes._subplots.AxesSubplot at 0x7f30f2ed4fd0>

Exercise 1.2

Install seaborn:

conda install seaborn

Use the Countplot to show the frequencies of thefts at the single location (descriptions).

Plot the top 10 hot spots horizontally.

See also Plotting with categorical data for further information about how to show categorical data with seaborn.

Hint

you still have to use value_counts() to find the top10 hot spots
set sns.countplot-order parameter to the top10 hot spots



In [47]:

    
import seaborn as sns
# TODO your code goes here

Exercise 2

Investigate, at which weekdays/which hour the most thefts happen.

Exercise 2.1

Convert Date to pandas.DateTime,
set Date as index and
add a further column Weekday (which is the weekday of Date), which is an ordered category (['Monday' , 'Tuesday'...).
Check
- the Pandas Docs
- and the Advanced Pandas Primer on how to convert to an ordered category.

Note:
Defining the weekday as categorical is required to get the plots in the further exercises ordered by weekday.



In [45]:

    
# TODO Date to pandas.DateTime



In [8]:

    
# TODO Date as index



In [9]:

    
# TODO WeekDay as new ordered categorical column

Exercise 2.2

Count the thefts by weekday and
make a corresponding histogram (barchart)

At which weekday happen the most thefts?



In [44]:

    
# TODO group by Weekday



In [18]:

    
# Plot group (barh)

Exercise 2.3

Add a further column Hour (which is the hour of Date) to the mvt-dataframe.
make a line-plot - 1 line per weekday - of number of thefts (y-axis) over the Hours (x-axis).
The plot should look similar to:

Note
To be able to plot without looping, you may want to

groupby(['Weekday','Hour']) and aggregate size
reset_index of the aggregate and pivot or unstack, so that index are the 24 Hours, the columns are the Weekdays and the values are the counts. Please check the
- Pandas pivot Documentation on how pivot works or
- Pandas unstack Documentation on how unstack works.
The pivot-table should look similar to:

Weekday	Monday	Tuesday	Wednesday	Thursday	Friday	Saturday	Sunday
Hour
0	76	..	..	..	..	..	234
1	126	..	..	..	..	..	...



In [43]:

    
# TODO add additional column Hour



In [42]:

    
# TODO group by Weekday,Hour and aggregate size
# TODO reset index (for pivot only)
# TODO pivot (or unstack)



In [41]:

    
# make a line-plot: count over Hour by Weekday(==Groups)

Exercise 2.4

Draw a heatmap with seaborn.heatmap:
x-axis=Hours, y-axis=Weekday, fill=count

At which weekday/time are the most thefts committed?

Additional
Play with color palettes for parameter cmap



In [40]:

    
# TODO

	Date	Block	Primary Type	Description	Location Description	Ward	Community Area	Year	Location
0	05/03/2016 08:00:00 PM	100XX S SANGAMON ST	MOTOR VEHICLE THEFT	AUTOMOBILE	STREET	34.0	73.0	2016	(41.711843569, -87.646607932)
1	05/03/2016 11:00:00 PM	084XX S MORGAN ST	MOTOR VEHICLE THEFT	AUTOMOBILE	STREET	21.0	71.0	2016	(41.740895923, -87.648617881)
2	05/03/2016 04:45:00 PM	003XX W MONROE ST	MOTOR VEHICLE THEFT	AUTOMOBILE	STREET	2.0	32.0	2016	(41.88063228, -87.635935494)