Pandas · Time series

I ended the Pandas phase with dates.

• pd.to_datetime parses dates. pd.date_range(start, periods, freq) generates them.
• Setting the dates as the index (a DatetimeIndex) unlocks slicing like df.loc["2024-03"].
• The .dt accessor gives the parts of a date: .dt.day_name(), .dt.month.
• resample("W") / resample("MS") buckets rows by week or month start. Think of it as groupby for time.
• rolling(7).mean() gives a moving average, and shift(1) gives the previous value (for day-over-day change).

import numpy as np
import pandas as pd

rng = np.random.default_rng(0)
idx = pd.date_range("2024-01-01", periods=60, freq="D")
df = pd.DataFrame({"steps": rng.integers(3000, 12000, size=60)}, index=idx)

print(df.loc["2024-02-01":"2024-02-03"])
print(df.resample("MS")["steps"].agg(["mean", "max"]).round(0))
df["ma7"] = df["steps"].rolling(7).mean().round(0)
df["change"] = df["steps"] - df["steps"].shift(1)
print(df.iloc[5:9])
print(df.index[:3].day_name().tolist())
print(df.groupby(df.index.day_of_week >= 5)["steps"].mean().round(0).rename({False: "weekday", True: "weekend"}))

Output:

            steps
2024-02-01   4580
2024-02-02   3803
2024-02-03  10768
              mean    max
2024-01-01  7845.0  11736
2024-02-01  7133.0  11974
            steps     ma7  change
2024-01-06   3368     NaN -2402.0
2024-01-07   3677  6461.0   309.0
2024-01-08   3148  5389.0  -529.0
2024-01-09   4577  4795.0  1429.0
['Monday', 'Tuesday', 'Wednesday']
weekday    7699.0
weekend    6956.0
Name: steps, dtype: float64