Plotting Grouped Datetime

Question

This post is sort of long, so here's the ultimate "ask" upfront:

Is there a way to transform the x-axis/index of the resulting groupby or a way to pass other types of arguments to the axvspan function?

I have a DataFrame with a datetime column, which I've grouped by year and weekofyear. This works okay, but the x-axis is displayed as a tuple. I want to axvspan, but I don't know who to deal with the tuples.

import numpy as np
import pandas as pd
import datetime
from matplotlib import pylab
import matplotlib.pyplot as plt
%matplotlib inline

query = ("https://data.cityofchicago.org/resource/6zsd-86xi.json?$where=year>2010")
raw_data = pd.read_json(query)

Here's an overview of the DataFrame. I'm going to be working with the date column.

raw_data.info()
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1706960 entries, 0 to 1706959
Data columns (total 22 columns):
arrest                  bool
beat                    int64
block                   object
case_number             object
community_area          float64
date                    datetime64[ns]
description             object
district                float64
domestic                bool
fbi_code                object
id                      int64
iucr                    object
latitude                float64
location                object
location_description    object
longitude               float64
primary_type            object
updated_on              object
ward                    float64
x_coordinate            float64
y_coordinate            float64
year                    int64
dtypes: bool(2), datetime64[ns](1), float64(7), int64(3), object(9)
memory usage: 263.7+ MB

First, filter all crimes but HOMICIDES.

# get murders
raw_data = raw_data[raw_data["primary_type"] == "HOMICIDE"]

# plot murder count by year and week of the year
plt.figure(figsize=(18, 6))
raw_data.groupby([raw_data.date.dt.year, 
                  raw_data.date.dt.weekofyear])["community_area"].size().plot()

So, as you can see, the x-axis is represented as tuples. Like I said before, I'd like to add a axvspan to add an arbitrary green span to my plot. If the x-axis maintained it's datetime structure, I could put values in the function like so, and it would work:

pylab.axvspan(datetime.strptime('2015-12-1 13:40:00', "%Y-%m-%d %H:%M:%S"), 
              datetime.strptime('2016-1-1 13:40:00', "%Y-%m-%d %H:%M:%S"), 
              facecolor='g', alpha=0.05) # green span 

This would shade the graph from December 1, 2015 to January 1, 2016 in green. Is there a way to transform the x-axis/index of the resulting groupby or a way to pass other types of arguments to the axvspan function?

Answer 1

Okay, I dusted off the ole Python for Data Analysis copy and re-discovered the resample method, and how well pandas handles time series data in general. The code below did the trick (sticking with my original data set):

# doesn't really matter which column I choose, I just picked one
murders = raw_data["community_area"]
murders.index = raw_data["date"]

plt.figure(figsize=(18, 6))
murders.resample("W-MON").count().plot() # weekly, every Monday

min_date = min(murders.index)
release_date = datetime.strptime('2015-11-24 12:00:00', "%Y-%m-%d %H:%M:%S") 
max_date = max(murders.index)

pylab.axvspan(min_date, 
              release_date, 
              facecolor='g', alpha=0.05) # green span
pylab.axvspan(release_date, 
              max_date, 
              facecolor='r', alpha=0.075) # red span
pylab.show()