In [2]:

import datetime
import numpy as np
import pandas as pd
print(pd.__version__)

1.3.5

In [156]:

start = datetime.datetime(2020, 1, 1)
end = datetime.datetime(2022, 1, 1)
# Create a DatetimeIndex with freq='Q'
# the 'Q' indicates the last day of each quarter
ts = pd.date_range(start, end, freq='M')

In [157]:

df = pd.DataFrame(index=ts.to_period('M'), data={'Quarter':ts.to_period('Q')})

In [158]:

# make a sine wave as data
df['Data'] = np.sin(np.linspace(0,10, len(ts)))

In [159]:

df['pct'] = df.groupby(by='Quarter')['Data'].transform(lambda x: x/sum(x))
df['cumpct'] = df.groupby(by='Quarter')['Data'].transform(lambda x: np.cumsum(x)/sum(x))
df['qtrTot'] = df.groupby(by='Quarter')['Data'].transform(lambda x: x.sum())

In [160]:

df.head()

Out[160]:

	Quarter	Data	pct	cumpct	qtrTot
2020-01	2020Q1	0.000000	0.000000	0.000000	1.185262
2020-02	2020Q1	0.421213	0.355376	0.355376	1.185262
2020-03	2020Q1	0.764049	0.644624	1.000000	1.185262
2020-04	2020Q2	0.964712	0.347751	0.347751	2.774149
2020-05	2020Q2	0.985865	0.355376	0.703126	2.774149

In [161]:

# We can do a shift of 3 months for our cumulative percent for the previous quarter
df['prev_qtr_pct'] = df.cumpct.shift(3)
# then divide it to apply the prediction
df['prev_qtr_linearity'] = df['Data'] / df['prev_qtr_pct']

In [162]:

# calculate the delta
df['delta'] = (df['prev_qtr_linearity']/df['qtrTot']) -1
# filter on month1/month2
scores = df.query('cumpct!=1 & delta.notna() & delta <=10000').delta

In [163]:

# MAPE:
np.abs(scores).mean()

Out[163]:

1.1399068142684166

In [164]:

np.abs(scores).plot(kind='bar')

Out[164]:

<AxesSubplot:>

In [67]:

df.groupby(by='Quarter').sum()['Data'].plot()

Out[67]:

<AxesSubplot:xlabel='Quarter'>

In [ ]: