Intermediate 15 min

Dates, Plots and Exporting

Work with time series, make quick charts and save results.

What you will learn

Parse dates
Resample by period
Plot and export

Business data is full of dates: sales per day, signups per week, server load per minute. pandas has first-class time support, and quick plotting turns numbers into pictures you can share.

import pandas as pd
import numpy as np

rng = pd.date_range("2026-01-01", periods=90, freq="D")
sales = pd.DataFrame({"revenue": np.random.default_rng(1).integers(100, 500, size=90)}, index=rng)

sales.index.dayofweek        # 0=Monday ... 6=Sunday
sales.index.month
sales.loc["2026-02"]         # partial-string selection: all of February
sales.loc["2026-01-15":"2026-01-20"]

Resampling

resample is groupby for time: regroup daily data into weeks or months.

weekly = sales["revenue"].resample("W").sum()
monthly = sales["revenue"].resample("MS").agg(["sum", "mean"])
print(monthly)

sales["7d_avg"] = sales["revenue"].rolling(7).mean()      # smooth out noise
sales["pct_change"] = sales["revenue"].pct_change()
sales["cumulative"] = sales["revenue"].cumsum()
sales["lag_1"] = sales["revenue"].shift(1)                # yesterday's value

df = pd.read_csv("orders.csv", parse_dates=["created"])
df["created"] = pd.to_datetime(df["created"], format="%d/%m/%Y")   # be explicit for ambiguous formats
df["weekday"] = df["created"].dt.day_name()
df["days_since"] = (pd.Timestamp.today() - df["created"]).dt.days

Store timestamps in UTC and convert for display with tz_localize and tz_convert when time zones matter.

pip install matplotlib

import matplotlib.pyplot as plt

ax = sales[["revenue", "7d_avg"]].plot(figsize=(9, 4), title="Daily revenue")
ax.set_ylabel("USD")
plt.tight_layout()
plt.savefig("revenue.png", dpi=150)

sales["revenue"].plot.hist(bins=15)                    # distribution
employees.groupby("country")["salary"].mean().plot.bar()   # category comparison

Line charts for trends over time; bar charts for comparing categories; histograms for distributions; scatter plots for relationships.
Always label axes and add a title. Start bar charts at zero.

monthly.to_csv("monthly.csv")
monthly.to_excel("monthly.xlsx")
df.to_parquet("orders.parquet")     # compact and fast for big data
df.to_sql("orders", engine, if_exists="replace", index=False)

Big files

For data that does not fit in memory, read in chunks (chunksize=), keep only the columns you need (usecols=), or move to Polars or DuckDB.

Try it yourself

Create 60 days of random data, compute a 7-day rolling mean and the best week, and save a line chart to a PNG file.

Show solution

idx = pd.date_range("2026-01-01", periods=60)
s = pd.Series(np.random.default_rng(0).normal(100, 15, 60), index=idx)
print(s.resample("W").sum().idxmax())
ax = pd.DataFrame({"value": s, "7d": s.rolling(7).mean()}).plot(title="Demo")
ax.figure.savefig("demo.png")