Pandas DataFrame-将月份转换为日期时间，并从多个列中反复选择数据进行绘图

Question

Say I have a pandas DataFrame with the format:

     Month Thing1 Thing2       Tot
0   Jan-12      A      Z  0.005880
1   Jan-12      A      Z  0.024500
...
20  Jan-12      B      Y  0.001533
21  Jan-12      C      X  0.003892
22  Jan-12      C      X  0.001680
23  Jan-12      C      X  0.001680
24  Jan-12      C      X  0.001680
25  Jan-12      C      X  0.001680
26  Jan-12      A      W  0.001680
27  Jan-12      D      V  0.013440
28  Jan-12      E      U  0.001680
...

The Month column goes unitl Apr-14. I am trying to plot line graphs for the monthly totals for each item in Thing1 and Thing2 .

I am attempting this using groupby :

a=pd.read_csv('all2.csv')
sums=a.groupby([u'Month',u'Thing1',u'Thing2']).sum()

which gives me:

Apr-12 A      W         6.427773
              Z         4.347471
       B      T         7.062425
              Y        17.183562
       C      X        14.583337
       D      V         0.114450
       E      U         0.008050
       F      Q         0.000490
              R         0.004468
       G      P         0.010932
       ...

However the months come up alphabetically. My questions are:

How can I get Pandas to consider the month column as a datetime object?

How can I iterate through Thing1 column and plot time series monthly totals for each item in Thing2 ?

I imagine there is a way to reorganise the Dataframe such that a simple call to plot() will do the job?

Answer 1

This is because your 'Month' column is not in the right dtype . You can get the intended result by firstly converting the Month column to datetime format:

df['Month']=pd.to_datetime(df.Month) , before calling df.groupby([u'Month',u'Thing1',u'Thing2']).sum()

But careful, Pandas doesn't know whether Jan-12 means 2014-01-12 or 2012-01 , by default it convert you data to the former. To get the latter, supply .to_datetime with format='%b-%y' argument.

For your second question, you can get the level of Thing1 by dfgb.index.get_level_values(1) . where dfgb is the DataFrame from groupby . Then you can plot the time series by:

for item in dfgb.index.get_level_values(1):
    dfgb.xs(item, level=1).plot(kind='bar') #for bar graph