Pandas - 将每个 groupby 系列与 Date 列进行细分，其中每个组中的行数不同

Question

I have a csv file which I have read into a Pandas Dataframe. The dataframe (say 'cdata') has the below columns

I want to be able to group this data by State and subplot the cumulative confirmed column data for each state in the same plot. The data will be plotted against the Date column.

The distribution of data against the Date column is not uniform ie not all State will have data row for each Date.

When I try to subplot this using the below the plotted data does not look okay.

fig,ax = plt.subplots(figsize=(8,6))
count=1;
for state,df in cdata.groupby('State'):
    if(count < 5):
        df.plot(x='Date', y='Confirmed', ax=ax, label=state)
        count = count +1

plt.legend()

This obviously does not look okay since if I look at the data the cumulative figure for State='Andhra Pradesh' on the 1st May is 1463 and not ~400 that the plotted graph seems to point.

What am I doing wrong here?

Answer 1

You are plotting the daily confirmed number and not the cumulative sum of confirmed. You can add a new column with the cumulative sum and plot it instead.

Also, be sure to set the 'Date' column as a date type and sort it before calculating the cumulative sum, you can do something like this:

## Transform 'Date' to datetime
df['Date'] = pd.to_datetime(df['Date'])

## Sort the df by the 'Date' column
df.sort_values('Date', inplace=True)

## Calculate cumulative sum of 'Confirmed' by state
df['Total Confirmed'] = df.groupby('State')['Confirmed'].transform('cumsum');

## Plot
fig,ax = plt.subplots(figsize=(8,6))
count=1
for state, df in cdata.groupby('State'):
    if(count < 5):
        df.plot(x='Date', y='Total Confirmed', ax=ax, label=state)
        count = count + 1

plt.legend()

Answer 2

I was able to achieve the outcome I was looking for with the below code. However I am sure this is not the most elegant way of achieving the same and am still looking for alternatives that are much more intuitive.

grouped = cdata.groupby(['Date','State'],sort=False)['Confirmed'].sum().unstack('State')
grouped.reset_index(inplace=True)
columns = grouped.columns.to_list()[1:-1]
fig,ax = plt.subplots(figsize=(20,14))
grouped.plot(x='Date',y=columns, ax=ax)