Working with Dates and Time Series

import pandas as pd
import numpy as np

# to_datetime — parse from string
dates = pd.to_datetime(['2024-01-15', '2024-02-20', '2024-03-10'])
print(dates)         # DatetimeIndex
print(dates.dtype)   # datetime64[ns]

# Various formats
pd.to_datetime('15-01-2024', format='%d-%m-%Y')
pd.to_datetime('January 15, 2024', format='%B %d, %Y')
pd.to_datetime('2024/01/15 14:30:00')

# In read_csv
df = pd.read_csv('sales.csv', parse_dates=['Date'])

# Create date range
date_range = pd.date_range(start='2024-01-01', end='2024-12-31', freq='D')  # Daily
months = pd.date_range('2024-01', periods=12, freq='MS')  # Month starts
business = pd.date_range('2024-01-01', periods=10, freq='B')  # Business days

df = pd.DataFrame({
    'Date': pd.date_range('2024-01-01', periods=365, freq='D'),
    'Revenue': np.random.randint(1000, 10000, 365)
})

# Extract date components
df['Year']    = df['Date'].dt.year
df['Month']   = df['Date'].dt.month
df['Day']     = df['Date'].dt.day
df['DayName'] = df['Date'].dt.day_name()
df['Quarter'] = df['Date'].dt.quarter
df['Week']    = df['Date'].dt.isocalendar().week
df['DayOfWeek'] = df['Date'].dt.dayofweek  # 0=Monday, 6=Sunday
df['IsWeekend'] = df['DayOfWeek'].isin([5, 6])

# Month name
df['MonthName'] = df['Date'].dt.month_name()

print(df[['Date', 'Year', 'Month', 'DayName', 'Quarter', 'IsWeekend']].head(10))

df = df.set_index('Date')  # Set datetime as index for easy filtering

# Filter by date range
jan_data = df['2024-01']                      # All of January
q1_data  = df['2024-01':'2024-03']            # Q1
week_data = df['2024-01-08':'2024-01-14']     # Specific week

# Filter using boolean
weekdays = df[~df['IsWeekend']]
high_revenue = df[df['Revenue'] > 8000]

# Filter by month/year
feb_data = df[df.index.month == 2]
recent = df[df.index >= '2024-06-01']

# resample — group time series by time frequency
df_reset = df.reset_index()

# Weekly summary
weekly = df['Revenue'].resample('W').agg(['sum', 'mean', 'max'])
print("Weekly Summary:\n", weekly.head())

# Monthly summary
monthly = df['Revenue'].resample('ME').agg(
    Total=('Revenue', 'sum'),      # ME = Month End
    Average=('Revenue', 'mean'),
    Peak=('Revenue', 'max')
) if False else df['Revenue'].resample('M').agg({'sum', 'mean', 'max'})

# Common resample frequencies:
# 'D'=Day, 'W'=Week, 'ME'=MonthEnd, 'QE'=QuarterEnd, 'YE'=YearEnd
# 'H'=Hour, 'T'=Minute, 'S'=Second

# Quarterly totals
quarterly = df['Revenue'].resample('Q').sum()
print("Quarterly Revenue:\n", quarterly)

# rolling() — moving window calculations
df['Revenue_7D_MA'] = df['Revenue'].rolling(window=7).mean()   # 7-day moving avg
df['Revenue_30D_MA'] = df['Revenue'].rolling(window=30).mean() # 30-day moving avg
df['Revenue_7D_Std'] = df['Revenue'].rolling(window=7).std()   # 7-day volatility

# Expanding window (cumulative from start)
df['Revenue_Cumsum'] = df['Revenue'].expanding().sum()
df['Revenue_CumMax'] = df['Revenue'].expanding().max()

print(df[['Revenue', 'Revenue_7D_MA', 'Revenue_30D_MA']].head(10))

# Time-based shifting
df['Revenue_Yesterday'] = df['Revenue'].shift(1)   # Previous day
df['Revenue_NextWeek']  = df['Revenue'].shift(-7)  # 7 days ahead
df['Revenue_Change']    = df['Revenue'].diff()     # Day-over-day change
df['Revenue_PctChange'] = df['Revenue'].pct_change() * 100  # % change