Plot multiple timeseries in one plot

Question

I would like to plot multiple timeseries (one for each value in de column cat) in one plot but haven't worked to ho to do that. My code so far is:

import numpy as np
import pandas as pd

dat = pd.date_range(start='1/1/2018', end='31/12/2018', freq='H')
num = ['A' + str(x).zfill(4) for x in range(len(dat))]
cat = np.random.choice(['A', 'B', 'C', 'D'], len(dat))

df = pd.DataFrame({'date': dat, 'num': num, 'cat':cat}).set_index('date')

print(df.groupby([pd.Grouper(freq='D'), 'cat']).count().unstack().fillna(0).astype(int))

Result:

           num            
cat          A   B   C   D
date                      
2018-01-01   7   3   5   9
2018-01-02   6   3   6   9
2018-01-03  11   3   8   2
2018-01-04   2   6   5  11
2018-01-05   4   8   4   8
2018-01-06   8   8   3   5
2018-01-07   5   8   6   5
2018-01-08   3   8   5   8

I would like to plot different combinations of categories (cat) like (A and B together or C and D together) in one timeseries plot with matplotlib or seaborn but are 'stuck' in de multilevelindexes...

Any suggestions how to select different combinations of columns and plot them? Maybe there is a better way than to unstack the data.

Answer 1

Yes, better is avoid MultiIndex in columns:

df1 = df.groupby([pd.Grouper(freq='D'), 'cat'])['num'].count().unstack(fill_value=0)

Or:

df1 = df.groupby([pd.Grouper(freq='D'), 'cat']).size().unstack(fill_value=0)

Then plot:

df1[['A','B']].plot()