Matplotlib — Pie & Scatter

• Draw a labelled pie chart with ax.pie.
• Recognise when a bar chart would be more honest.
• Draw a scatter plot with ax.scatter.
• Add a trend line and read correlation visually.

pie     proportions of a single whole, ≤ 5 slices
bar     comparing categories — almost always better than pie
line    change over time
scatter relationship between TWO numeric variables

Pie chart

import matplotlib.pyplot as plt

labels = ["Roti", "Milo", "Nasi", "Teh"]
counts = [120, 95, 180, 60]

fig, ax = plt.subplots(figsize=(6, 6))
ax.pie(counts, labels=labels, autopct="%1.1f%%",
       startangle=90, counterclock=False)
ax.set_title("Sales share · May 2026")
fig.savefig("share.png", dpi=200)

• autopct="%1.1f%%" prints each slice's percentage on it.
• startangle=90, counterclock=False = the canonical "top, clockwise" order.
• An explode tuple separates one slice for emphasis.

Scatter

x = df["hours_studied"]
y = df["score"]

fig, ax = plt.subplots(figsize=(7, 5))
ax.scatter(x, y, alpha=0.6, s=40)
ax.set_xlabel("Hours studied")
ax.set_ylabel("Score")
ax.set_title("Study time vs score")

alpha makes overlapping dots visible; s is the marker size.

Trend line

import numpy as np
m, b = np.polyfit(x, y, 1)        # linear fit
ax.plot(x, m*x + b, color="red", linewidth=2, label=f"y = {m:.2f}x + {b:.1f}")
ax.legend()

Encode a third variable

ax.scatter(x, y, c=df["age"], cmap="viridis", s=df["score"])
# colour by age, size by score

import pandas as pd
import matplotlib.pyplot as plt
import numpy as np

df = pd.read_csv("clean.csv", parse_dates=["date"])
df["total"] = df["quantity"] * df["price"]

# Pie: revenue share by product
share = df.groupby("product")["total"].sum().sort_values(ascending=False)

fig, axes = plt.subplots(1, 2, figsize=(12, 5))

axes[0].pie(share, labels=share.index, autopct="%1.1f%%",
            startangle=90, counterclock=False,
            wedgeprops={"edgecolor": "white", "linewidth": 1})
axes[0].set_title("Revenue share by product")

# Scatter: quantity vs price (per order)
axes[1].scatter(df["quantity"], df["price"], alpha=0.6, s=40)
m, b = np.polyfit(df["quantity"], df["price"], 1)
xs = np.linspace(df["quantity"].min(), df["quantity"].max(), 50)
axes[1].plot(xs, m*xs + b, color="red", linewidth=2,
             label=f"trend: y = {m:.2f}x + {b:.1f}")
axes[1].set_xlabel("Quantity per order")
axes[1].set_ylabel("Price (RM)")
axes[1].set_title("Order size vs unit price")
axes[1].legend()

fig.tight_layout()
fig.savefig("share_and_corr.png", dpi=200)
plt.show()

Read the diff

Two panels, two stories. The pie answers "what fraction of revenue comes from each product?"; the scatter (with a fitted trend line) answers "do bigger orders come at lower per-unit prices?". The legend on the scatter shows the regression equation so the trend is auditable.

Pie charts: only for proportions of a whole, ≤ 5 slices. Otherwise use a bar. Scatter: two numerics; alpha + size to handle overlapping; fit a trend line with np.polyfit for clarity. Tomorrow we compose multiple charts on one figure.