Python for Data (pandas + NumPy)

After this lesson, you will be able to: Use pandas and NumPy for basic data manipulation, plot results with matplotlib, and know when to reach for them vs plain Python.

Pandas + NumPy are how Python eats data, and matplotlib is how you chart it. Just enough to be dangerous + know when to escape to them.

Prerequisites:Python Scripting

NumPy basics

Array-based numeric computing. 100x faster than Python lists for math.

python

import numpy as np

a = np.array([1, 2, 3, 4, 5])
print(a * 2)           # [2 4 6 8 10], vectorized; no loop
print(a.mean(), a.std(), a.sum())

# 2D array
m = np.array([[1, 2], [3, 4]])
print(m.shape)         # (2, 2)
print(m @ m)           # matrix multiply

# Random + seeding
rng = np.random.default_rng(seed=42)
samples = rng.normal(loc=0, scale=1, size=1000)

Pandas basics

DataFrame = a programmable spreadsheet.

python

import pandas as pd

# Load CSV
df = pd.read_csv("sales.csv", parse_dates=["date"])

# Inspect
df.head()
df.describe()
df.dtypes

# Filter + select
recent = df[df["date"] >= "2026-01-01"]
emails = df[["name", "email"]]

# Group + aggregate
by_region = df.groupby("region")["revenue"].agg(["sum", "mean", "count"])

# Join
orders = pd.read_csv("orders.csv")
merged = df.merge(orders, on="user_id", how="left")

# Write back
merged.to_csv("out.csv", index=False)
merged.to_parquet("out.parquet")    # binary, smaller, faster

Plotting with matplotlib

matplotlib is the baseline plotting library. pandas plots through it directly, so a quick chart is one method call. Keep this light; the AI/ML track goes deeper into visualization.

python

import matplotlib.pyplot as plt
import pandas as pd

df = pd.read_csv("sales.csv", parse_dates=["date"])

# pandas plots through matplotlib: one call for a quick line chart
by_month = df.groupby(df["date"].dt.to_period("M"))["revenue"].sum()
by_month.plot(kind="line", title="Revenue by month")
plt.ylabel("Revenue (USD)")
plt.tight_layout()
plt.savefig("revenue.png", dpi=150)   # save for a report
plt.show()                            # or display interactively

# Drop to the matplotlib API directly when you need control
fig, ax = plt.subplots(figsize=(8, 4))
ax.bar(by_region.index, by_region["sum"])
ax.set_title("Revenue by region")
ax.set_xlabel("Region")
ax.set_ylabel("Revenue (USD)")
fig.savefig("by_region.png")
# Common next steps: seaborn (prettier statistical charts on top of matplotlib),
# or plotly for interactive charts.

When to use pandas vs plain Python

Tabular data with > ~10K rows: pandas. Numeric arrays, math, ML pre-processing: NumPy. Quick charts from a DataFrame: df.plot(...) (matplotlib under the hood); save with plt.savefig for reports. Small CSV (< 1K rows), one-off scripts: csv + dataclasses might be cleaner. Streaming / large data: polars (pandas alternative, 10x faster) or DuckDB.

Common mistakes only experienced data devs avoid

Iterating with `for i, row in df.iterrows()`. SLOW. Vectorize with column ops. Using `apply()` to do what a vectorized op can. Floating CSV index column from one tool to another. Always to_csv(index=False) unless meaningful. Mixing dtypes (string in a numeric column). Pandas degrades performance.

Quick Check

When should you use pandas iterrows()?

Pick the honest answer.

AlwaysAlmost never, vectorized ops are 100x fasterWhen working with stringsWhen the file is big

←Python for Scripting and Automation

Back to Python (Professional)

Async Python (asyncio)→