Synchronisation

# Create a time series with a frequency of 1 hour and a length 72*24 hours
import pandas as pd
import numpy as np

rng = np.random.default_rng(1979)
stock = rng.standard_normal(72).cumsum() + 2
days = pd.date_range('2021-01-01', periods=72, freq='D')

rng2 = np.random.default_rng(1000)
electricity = rng2.standard_normal(72*24).cumsum()+30
hours = pd.date_range('2021-01-01', periods=72*24, freq='h')

# Plot the two series in the same plot
import matplotlib.pyplot as plt
fig, ax = plt.subplots(figsize=(7, 2))
ax.plot(days, stock, label='stock')
ax.plot(hours, electricity, label='electricity')
ax.legend()
plt.show()

# Turn the hourly series into a daily series by taking the mean of each day
daily2 = electricity.reshape(72, 24).mean(axis=1)
print(daily2.shape)
daily2[:5]

# Make a dataframe of days and daily
df_days = pd.DataFrame({'days': days, 'stock': stock})
# .. and hours and hourly
df_hours = pd.DataFrame({'hours': hours, 'electricity': electricity})
df_hours.head()

# Group the hourly series by day and take the mean of each day
daily3 = df_hours.electricity.groupby(df_hours.hours.dt.date).mean()
daily3.name = 'electricity' # <- Will be column name in dataframe

# Change name of index to days
daily3.index.name = 'days'

# Convert index to datetime
daily3.index = pd.to_datetime(daily3.index) # <- Important for matching with other datetimes!
daily3.head()

# Concatenate the daily series and the daily3 series
daily4 = pd.concat([pd.DataFrame({"stock":stock}, index=days), daily3], axis=1)
daily4

# Plot the two time series in daily4 using Pandas plot
daily4.plot()
plt.show()

# Create a time series with irregular frequency
irr_dates = pd.date_range('2000-01-01 00:00', periods=365, freq='D') # <- Higher accuracy applied here to enable sampling

# Sample 50 random dates from irr_dates
m = np.arange(0,365,1)
rng.shuffle(m)
irr_dates = irr_dates[np.sort(m[:50])].sort_values()

# Create a series with random values and the sampled dates as index
irr_series = pd.DataFrame({'values': rng.standard_normal(50)}, index=irr_dates)
#irr_series.head()
print(irr_series)

# Interpolate the series to weekly frequency without intermediate resampling
weekly = irr_series.resample('W').interpolate("linear")
weekly.head()

# Plot the two series
fig, ax = plt.subplots(figsize=(7, 2))
ax.plot(irr_series, label='irregular')
ax.plot(weekly, 'o', label='weekly')
ax.legend()
plt.show()

# Interpolate the series to weekly frequency after resampling to daily frequency
daily_x = irr_series.resample('D').interpolate("linear")
weekly = irr_series.resample('D').interpolate("linear").resample('W').interpolate("linear")
weekly.head()

# Plot the two series
fig, ax = plt.subplots(figsize=(7, 2))
ax.plot(daily_x, 'o', label='interpolated daily')
ax.plot(irr_series, 'o', label='irregular')
ax.plot(weekly, label='weekly')
ax.legend()
plt.show()